239 82 8MB
English Pages 241 Year 2013
FRONTIERS IN ELECTRONICS Proceedings of the Workshop on Frontiers in Electronics 2009
8370.9789814383714-tp.indd 1
23/4/12 4:51 PM
SELECTED TOPICS IN ELECTRONICS AND SYSTEMS Editor-in-Chief:
M. S. Shur
Published* Vol. 40: SiC Materials and Devices — Vol. 1 eds. M. S. Shur, S. Rumyantsev and M. Levinshtein Vol. 41: Frontiers in Electronics Proceedings of the WOFE-04 eds. H Iwai, Y. Nishi, M. S. Shur and H. Wong Vol. 42: Transformational Science and Technology for the Current and Future Force eds. J. A. Parmentola, A. M. Rajendran, W. Bryzik, B. J. Walker, J. W. McCauley, J. Reifman, and N. M. Nasrabadi Vol. 43: SiC Materials and Devices — Vol. 2 eds. M. S. Shur, S. Rumyantsev and M. Levinshtein Vol. 44: Nanotubes and Nanowires ed. Peter J. Burke Vol. 45: Proceedings of the 2006 IEEE Lester Eastman Conference on Advanced Semiconductor Devices eds. Michael S. Shur, P. Maki and J. Kolodzey Vol. 46: Terahertz Science and Technology for Military and Security Applications eds. Dwight L. Woolard, James O. Jensen, R. Jennifer Hwu and Michael S. Shur Vol. 47: Physics and Modeling of Tera- and Nano-Devices eds. M. Ryzhii and V. Ryzhii Vol. 48: Spectral Sensing Research for Water Monitoring Applications and Frontier Science and Technology for Chemical, Biological and Radiological Defense eds. D. Woolard and J. Jensen Vol. 49: Spectral Sensing Research for Surface and Air Monitoring in Chemical, Biological and Radiological Defense and Security Applications eds. J.-M. Theriault and J. Jensen Vol. 50: Frontiers in Electronics eds. Sorin Cristoloveanu and Michael S. Shur Vol. 51: Advanced High Speed Devices eds. Michael S. Shur and P. Maki Vol. 52: Frontiers in Electronics Proceedings of the Workshop on Frontiers in Electronics 2009 eds. Sorin Cristoloveanu and Michael S. Shur
*The complete list of the published volumes in the series can be found at http://www.worldscibooks.com/series/stes_series.shtml
Selected Topics in Electronics and Systems – Vol. 52
FRONTIERS IN ELECTRONICS Proceedings of the Workshop on Frontiers in Electronics 2009 Rincon, Puerto-Rico
13 – 16 December 2009
Editors
Sorin Cristoloveanu Grenoble INP – Minatec, France
Michael S. Shur Rensselaer Polytechnic Institute, USA
World Scientific NEW JERSEY
8370.9789814383714-tp.indd 2
•
LONDON
•
SINGAPORE
•
BEIJING
•
SHANGHAI
•
HONG KONG
•
TA I P E I
•
CHENNAI
23/4/12 4:51 PM
Published by World Scientific Publishing Co. Pte. Ltd. 5 Toh Tuck Link, Singapore 596224 USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601 UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE
British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library.
Selected Topics in Electronics and Systems — Vol. 52 FRONTIERS IN ELECTRONICS Proceedings of the Workshop on Frontiers in Electronics 2009 Copyright © 2013 by World Scientific Publishing Co. Pte. Ltd. All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means, electronic or mechanical, including photocopying, recording or any information storage and retrieval system now known or to be invented, without written permission from the Publisher.
For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy is not required from the publisher.
ISBN 978-981-4383-71-4
Printed in Singapore
PREFACE The Workshop on Frontiers in Electronics – WOFE–09 – took place in Rincon, Puerto-Rico, in December of 2009. This meeting was the sixth in the series of the WOFE workshops (see Figure 1) and strongly reinforced the tradition of scientific quality and visionary research. Fifty leading experts from academia, industry, and government agencies came to Puerto Rico to report on the most recent and exciting developments in their fields. The warm and friendly atmosphere was ideal for inciting the participants to exchange frank and sometimes rather controversial views on future trends and directions of the electronics and photonics industry. The conference was sponsored by the National Science Foundation, Rensselaer Polytechnic Institute, and Seoul National University. Two companies – Nanoscience, Inc. and Sensor Electronic Technology, Inc. sponsored Best Paper Award, Best Student paper Award and Innovation Paper Award. IEEE EDS provided Technical Sponsorship. Mr. V. S. Chivukula of RPI received Best Student Paper Award for his paper entitled “Wide Band Gap Semiconductor LC Oscillator based Ultraviolet Sensors”. Mr. Wu of the University of Puerto Rico received Nanoscience Best Student Paper Award for the paper entitled “Room Temperature Ferromagnetic Behavior in Yb-doped GaN Semiconductor”. Best Poster Paper Award went to G. Belenky and colleaguess and the Nanoscience Best Student Poster Award was received by Mr. S. Ghosh. Professor Fernando A. Ponce received the SET innovation Award, while Nanoscience Best Paper Award went to Professor Enrique Calleja. Finally, Professor Kim M. Lewis received the Best Paper Award for the paper entitled ‘Molecular Conductance: Exploring Functionality in Simple Circuits’. This issue includes the best papers of WOFE-09 invited by the Editors and down selected after the peer review. The aim of this book was to collect the best papers of the conference, in an extended version, and make them known in the international arena. These papers are divided into four sections: advanced terahertz and photonics devices; silicon and germanium on insulator and advanced CMOS and MOSHFETs; nanomaterials v
vi
Preface
and nanodevices; and wide band gap technology for high power and UV photonics. These key issues are in forefront of the microelectronics research and many papers in this book go well beyond the discussions of the original work giving a good overview of the field. This book will be useful for nano-microelectronics scientists, engineers, and visionary research leaders. It is also recommended to graduate students working at the frontiers of the nanoelectronics and microscience. On behalf of the WOFE Organizing, Program, and Steering Committees, we would like to thank all participants and especially the invited contributors to this issue for making WOFE–09 a successful conference. We also gratefully acknowledge generous support of this workshop by the National Science Foundation. Our special thanks go to the Members of Organizing, Program, and Steering Committees, and to Session Organizers for their tireless work and inspiration. The next WOFE will take place in December 2011 somewhere in the Caribbean. We welcome suggestions for topics to be addressed, special sessions, and tutorials. We believe that nanoelectronics and microelectronics are the cross roads. WOFE-2011 will provide a unique opportunity to discuss what has to be done to stimulate NanoMicro-Electronics R&D during and after the on-going economic crisis. This material is based upon work supported by the National Science Foundation under Grant No. 0939894. Sorin Cristoloveanu and Michael Shur Editors
October 20, 2011 12:4 WSPC/120-IJHSES
02˙Contents
CONTENTS Preface
v
Advanced Terahertz and Photonics Devices
1
Broadband Terahertz Wave Generation, Detection and Coherent Control Using Terahertz Gas Photonics J. Liu, J. Dai, X. Lu, I. C. Ho and X. C. Zhang
3
How do We Lose Excitation in the Green? C. Wetzel, Y. Xia, W. Zhao, Y. Li, M. Zhu, S. You, L. Zhao, W. Hou, C. Stark and M. Dibiccari
13
Silicon Finfets as Detectors of Terahertz and Sub-Terahertz Radiation W. Stillman, C. Donais, S. Rumyantsev, M. Shur, D. Veksler, C. Hobbs, C. Smith, G. Bersuker, W. Taylor and R. Jammy
27
Progress in Development of Room Temperature CW GaSb based Diode Lasers for 2-3.5 µm Spectral Region T. Hosoda, J. Chen, G. Tsvid, D. Westerfeld, R. Liang, G. Kipshidze, L. Shterengas and G. Belenky WDM Demultiplexing by Using Surface Plasmon Polaritons D. K. Mynbaev and V. Sukharenko Silicon and Germanium on Insulator and Advanced CMOS and MOSHFETs
43
51
63
Connecting Electrical and Structural Dielectric Characteristics G. Bersuker, D. Veksler, C. D. Young, H. Park, W. Taylor, P. Kirsch, R. Jammy, L. Morassi, A. Padovani and L. Larcher
65
Advanced Solutions for Mobility Enhancement in SOI MOSFETs L. Pham-Nguyen, C. Fenouillet-Beranger, P. Perreau, S. Denorme, G. Ghibaudo, O. Faynot, T. Skotnicki, A. Ohata, M. Casse, I. Ionica, W. van den Daele, K-H. Park, S-J. Chang, Y-H. Bae, M. Bawedin and S. Cristoloveanu
81
Electron Scattering in Buried InGaAs/High-K MOS Channels S. Oktyabrsky, P. Nagaiah, V. Tokranov, M. Yakimov, R. Kambhampati, S. Koveshnikov, D. Veksler, N. Goel and G. Bersuker
95
vii
October 20, 2011 12:4 WSPC/120-IJHSES
viii
02˙Contents
Contents
Low Frequency Noise and Interface Density of Traps in InGaAs MOSFETs with GdScO3 High-K Dielectric S. Rumyantsev, W. Stillman, M. Shur, T. Heeg, D. G. Schlom, S. Koveshnikov, R. Kambhampati, V. Tokranov and S. Oktyabrsky
105
Low-Power Biomedical Signal Monitoring System for Implantable Sensor Applications M. R. Haider, J. Holleman, S. Mostafa and S. K. Islam
115
Nanomaterials and Nanodevices
129
III-V Compound Semiconductor Nanowires for Optoelectronic Device Applications Q. Gao, H. J. Joyce, S. Paiman, J. H. Kang, H. H. Tan, Y. Kim, L. M. Smith, H. E. Jackson, J. M. Yarrison-Rice, J. Zou and C. Jagadish
131
Electron Heating in Quantum-Dot Structures with Collective Potential Barriers L. H. Chien, A. Sergeev, N. Vagidov, V. Mitin and S. Birner
143
Electronic Structure of Graphene Nanoribbons Subjected to Twist and Nonuniform Strain A. Dobrinsky, A. Sadrzadeh, B. I. Yakobson and J. Xu
153
Low-Frequency Electronic Noise in Graphene Transistors: Comparison with Carbon Nanotubes G. Liu, W. Stillman, S. Rumyantsev, M. Shur and A. A. Balandin
161
ZnO Nanocrystalline High Performance Thin Film Transistors B. Bayraktaroglu, K. Leedy and R. Neidhard
171
Zinc Oxide Nanoparticles for Ultraviolet Photodetection S. Sawyer, L. Qin and C. Shing
183
Carbon-Based Nanoelectromechanical Devices S. Bengtsson, P. Enoksson, F. A. Ghavanini, K. Engstr¨ om, P. Lundgren, E. E. B. Campbell, J. Ek-Weis, N. Olofsson and A. Eriksson
195
Charge Puddles and Edge Effect in a Graphene Device as Studied by a Scanning Gate Microscope J. Chae, H. J. Yang, H. Baek, J. Ha, Y. Kuk, S. Y. Jung, Y. J. Song, N. B. Zhitenev, J. A. Stroscio, S. J. Woo and Y.-W. Son
205
April 3, 2013 14:55 WSPC/120-IJHSES
02˙Contents
Contents
Wide Band Gap Technology for High Power and UV Photonics Novel Approaches to Microwave Switching Devices using Nitride Technology G. Simin, J. Wang, B. Khan, J. Yang, A. Sattu, R. Gaska and M. Shur Author Index
ix
217 219
229
This page intentionally left blank
ADVANCED TERAHERTZ AND PHOTONICS DEVICES
This page intentionally left blank
BROADBAND TERAHERTZ WAVE GENERATION, DETECTION AND COHERENT CONTROL USING TERAHERTZ GAS PHOTONICS JINGLE LIU, JIANMING DAI, XIAOFEI LU, I-CHEN HO and X.-C. ZHANG Center for Terahertz Research, Rensselaer Polytechnic Institute, 110 8th Street, Troy, New York 12180, US [email protected] Terahertz (THz) gas photonics uses gas as THz emitter and sensor for time-domain spectroscopy. Unique properties of the gas promise scalable, strong THz wave generation with broad spectral range covering the entire THz gas (0.3 THz to 35 THz). The systematic study of THz wave generation and detection in different gases shows that the generation efficiency is monotonically decreasing with the ionization potential of the gas molecules while the detection efficiency is linearly proportional to the third order nonlinear coefficient of the gas molecules. We also discuss the development of THz wave detection using laser-induced fluorescence and coherent control with THz gas photonics. Keywords: Terahertz wave; spectroscopy; laser-induced plasma; fluorescence.
1. Introduction With the fast development of laser technologies in the past decades, ultrafast timedomain THz spectroscopic sensing and imaging have been widely available and used in areas of semiconductor characterization, security screening, industrial inspection and communications [1-3]. Among the various THz generation and detection methods, THz gas photonics has attracted a great amount of research interest and efforts in recent years due to its broad spectral coverage and scalability [4-9]. Unlike solids, gases do not have a damage threshold, or experience phonon absorption and interface reflection. This uniqueness makes gas an ideal emitter and sensor with continuous and broadband spectral response. The only limitation for the previously reported 10 THz bandwidth is the laser pulse duration [9]. Through using even shorter and stronger laser pulses, the THz spectral range using gas photonics can be further extended to 35 THz (1% of the spectrum maximum) and the peak field can reach up to 100 kV/cm. These advances will be driving many new discoveries in fundamental molecular, biological and materials dynamics. In this paper, we will cover the state-of-art technologies in THz gas photonics, including THz air generation, detection and coherent manipulation. 2. The Mechanism of THz Generation and Detection in Gas THz wave generation in gases has emerged as one of most promising broadband and high-intensity THz techniques. There are several ways to generate THz waves in gases such as single-color excitation, dual-color excitation and DC-biased excitation [10]. Here we focus our discussion on the THz generation scheme using dual-color laser pulse excitation which is most commonly used due to its high efficiency and controllability. In 3
4
J. Liu et al.
the process of THz wave generation in ambient air, one fundamental laser pulse ( ) and its second harmonic pulse (2 ) are collinearly focused into air. The second harmonic pulse is typically generated via frequency doubling by passing the fundamental pulse through a nonlinear beta barium borate crystal (BBO). The typical total optical intensity at the focus is roughly 1013 ~ 1014W/cm2 or even higher. At the optical focus, the strong laser field will release electrons from the air molecules through multi-photon or tunnel ionization. As a result, intense THz pulses are emitted from the ionized air. Four-wavemixing theory [4] and plasma transient current model [8] have been proposed to explain the physical process of THz air generation. The reciprocal, nonlinear optical process for THz wave generation can be used to detect THz wave in gases. Dai et al demonstrated that a THz pulse can be detected coherently in air by using a relatively strong probe pulse [7]. Since air has to be broken down for coherent detection, this method is called THz air-breakdown coherent detection (THz-ABCD). Later Karpowicz et al introduced a new coherent THz detection method using an external AC bias on the probe plasma as a local oscillator [9]. This heterodyne method is called THz air-biased-coherent-detection (also abbreviated as THz-ABCD). By referencing the modulation frequency of the AC bias to the lock-in amplifier, the measured second harmonic intensity is given by
I 2ω ∝ χ (3) Iω
2
Ebias ETHz
(1)
Where the χ (3) is the third-order nonlinear optical coefficient of the gas and Ebias is the applied AC bias. Since bias-induced second harmonic intensity can be used as a local oscillator, the coherent THz detection can be realized without using the high probe energy. The typically used pulse energy is ~100 µJ, depending on the probe pulse duration and gas ionization potential. This allows one to utilize most of the laser power for the generation part to achieve a high THz field.
Fig. 1. The schematics of the experimental setup of broadband THz generation and detection using gas. HV, high voltage. BBO, beta barium borate. Filter, 400 nm band-pass filter. Detector, photo-multiplier tube.
Broadband THz Wave Generation, Detection and Coherent Control Using THz Gas Photonics
5
Fig. 1 illustrates the typical experimental set-up of the THz air generation and air detection. Laser pulse with energy of sub mJ to a few mJ is used for THz air generation. The probe pulse is focused at the THz beam focal point. A 20 kV/cm high voltage AC bias is applied on the plasma formed by the probe pulse. The second harmonic photons measured by the photo-multiplier tube, are composed of the second harmonic photons generated through FWM of fundamental beam and THz beam, and that generated through FWM of fundamental beam and the bias field.
Fig. 2. (a) Typical time-domain waveform of a THz pulse using air generation and air detection. A 35 fs short laser pulse is used for generation and detection. (b) The corresponding THz spectrum in the frequency domain.
Fig. 2(a) shows a typical waveform of the THz pulse using air generation and air detection with bias. A laser pulse with 35 fs pulse duration, 600 µJ pulse energy and 800 nm center wavelength is used here. The peak THz field is about 70 kV/cm. The corresponding THz spectrum is plotted in Fig. 2(b). Due to the shorter pulse duration, the THz spectrum covers up to 35 THz (1% of the spectrum maximum). The small dip around 18.5 THz in the spectrum is due to the THz absorption in the silicon filter.
6
J. Liu et al.
3. THz Wave Generation and Detection in Different Gases In addition to the study of ambient air and dry nitrogen, the THz wave generation and detection has been investigated in other gases such as alkane gases and noble gases to reveal the role of the gas species in the efficiency of THz wave generation. The experiment of THz generation in different gases was performed in a supersonic gas jet where the background pressure was kept below 10-5 torr. The back pressure of the gas jet is 50 torr for all the gases. The reason that a gas jet is used rather than a gas cell, is that the intensity clamping [11], self-focusing or other nonlinear optical effects during the laser propagation in gas media can be significantly reduced so that the measured THz wave generation efficiency only depends on the gas species. Electro-optical (EO) sampling is used for the measurement of the time-domain THz waveform [12]. The dependence of the peak THz field on the ionization potential of gases was measured and the results are plotted in Fig. 3.
Fig. 3. Peak field amplitude of the THz pulse generated from different gases versus the ionization potential.
The results indicate that the THz wave generation efficiency monotonically decreases as the ionization potential of the gas increases. At the same gas pressure and laser intensity, the THz generation efficiency in n-butane (C4H10) whose ionization potential is 10.53 eV, is two hundred times larger than that in He whose ionization potential is 24.59 eV. The higher THz generation efficiency in the gas of lower ionization potential is attributed to the higher nonlinear coefficient (3) in the four-wave-mixing model or larger plasma density/current in the plasma transient current model. Broadband THz wave detection has been systematically investigated using selected gases. The dependencies of the detected second harmonic intensity on the third order nonlinear susceptibility are studied with nitrogen and alkane gases. The results are shown as blue dots in Fig. 4. The gas pressure is 100 torr for all the gases. For the sample in liquid phase, a gas dilute system is used to obtain pure vapor. The probe pulse energy is 15 µJ. A low bias field of 5 kV/cm is used to eliminate the electrical breakdown. The blue dashed line is a quadratic fitting of the data. The agreement between the
Broadband THz Wave Generation, Detection and Coherent Control Using THz Gas Photonics
7
measurement and the fitting further confirms Eq. (1). It has been found that C6H14 generates a second harmonic signal nearly two orders higher than nitrogen.
Fig. 4. The measured THz-induced second harmonic generation from laser plasma as a function of the thirdorder nonlinear coefficient (3). [13]
4. THz Remote Generation and Detection THz spectroscopy at remote distance has attracted a great amount of interest and effort due to its promising applications in areas of homeland security, environmental science and aerospace technologies. However, THz remote sensing has been a big challenge because of strong, ambient moisture absorption in the THz frequency range. Conventional THz generation and detection techniques cannot be applied to remote THz spectroscopy without suffering substantial attenuation during the THz wave propagation in the air. THz wave gas photonics provides a unique opportunity to overcome this longexisting barrier.
Fig. 5. THz waveforms measured at different distances. THz generation with two-color excitation is used and a 1 mm thick ZnTe crystal is used to measure the THz waveform.
In order to minimize the THz wave propagation loss in the air, plasma is created as the THz emitter by focusing two-color pulses at standoff distances [14]. The geometry of
8
J. Liu et al.
the focusing optics, the optical chirp and relative delay between two pulses are parameters which can be tuned for the optimization of THz generation. Fig. 5 shows the time-domain THz waveforms measured by EO sampling at distances of < 0.5 m, 6.5 m and 14 m, respectively. It can be seen that by precisely controlling the focusing of the two-color pulses and their relative phase, THz pulse generation can be realized at standoff distance. In this demonstration, total pulse energy of less than 1 mJ is used for THz generation. With use of the higher-power laser, the maximal distance at which a THz wave can be generated is expected to be further extended to hundreds of meters. With regards to THz remote detection, recently Liu et al introduced an omnidirectional coherent THz detection technique using radiation-enhanced-emission-offluorescence (REEF) from gas plasma [15]. It was experimentally found and theoretically explained that the interaction between the laser-induced plasma and THz pulses leads to the enhancement of the fluorescence emission. The enhanced fluorescence depends on the THz waveform, THz field amplitude and time delay between THz pulse and plasmagenerating laser pulse.
Fig. 6. (a) The nitrogen fluorescence spectra in THz fields of different strength. (b) The dependence of the fluorescence line emission at 357 nm and quadratic fit. Inset, the angular distribution of the fluorescence emission.
Broadband THz Wave Generation, Detection and Coherent Control Using THz Gas Photonics
9
When the laser beam is focused into the air, the strong laser field excites or ionizes the nitrogen and oxygen molecules. The strongest fluorescence band of the nitrogen molecule lies in the spectral range of 300 nm and 400 nm. The nitrogen fluorescence spectra are measured under the influence of the THz field strengths of 0, 75 and 100 kV/cm, respectively and the results are shown in Fig. 6(a). All line emission from the second positive band (2P) and first negative band (1N) of nitrogen molecules are increased homogeneously by the THz field. The measured field dependence of the line emission at 357 nm and quadratic fitting are plotted in Fig. 6(b). The inset shows the isotropic angular distribution of the fluorescence emission. This omni-directional emission pattern allows one to measure THz waves in the backward direction, which is crucial to the remote THz wave sensing. The mechanism for THz-REEF can be explained by the THz-wave-induced electron heating, electron-impact excitation/ionization of the high-lying Rydberg states and the subsequent increase of the fluorescing upper states. Specifically the electrons gain more kinetic energy from the THz field and then transfer part of the energy to the neighboring air molecules by collision. The more energy the electrons lose to molecules, the more upper states that can be excited. The linearity of this energy transfer process agrees well with measured quadratic dependence of the enhanced fluorescence on the THz field applied on the plasma.
Fig. 7. The vector potential, THz waveform measured by REEF and THz waveform measured by EO, respectively. Inset shows the corresponding THz spectra by REEF and EO.
Liu et al also demonstrated coherent THz detection by introducing a local oscillator to the THz-plasma interaction. A 20 kV/cm AC bias with the same direction as the THz field is applied on the plasma. The AC bias is set to 500 Hz, half of the laser repetition rate, similar to THz-ABCD. The vector potential of the THz pulse can be directly measured by lock-in amplifier which is referenced to AC bias frequency. The THz field can be obtained by performing the derivative of the vector potential. Fig. 7 shows the
10
J. Liu et al.
THz waveform measured by REEF, as described above, compared to that measured by the EO sampling. The inset shows the corresponding spectra of the two methods. The agreement between the results from the two methods indicates that THz-REEF is able to perform THz spectroscopic measurements. Further efforts could be done toward replacing the bias with the optical bias to achieve remote sensing without putting anything close to the target. By combining the THz air generation and THz-REEF, a broadband THz spectrometer with remote capability can be realized. This optical, remote THz spectroscopy is a promising and powerful means of imaging in the areas of the homeland security, environmental monitoring, etc. In addition to the application in remote sensing, THz-REEF can also be utilized in other laser plasma applications such as plasma diagnostics [16]. 5. Coherent Control of THz Generation The trajectory of the electrons ionized from the gas atoms or molecules by circularly or elliptically polarized femtosecond and 2 pulses has been systematically investigated independently by two groups [17, 18]. Both groups found similar results that THz polarization rotates as the relative phase between two pulses is tuned. This provides a new approach of coherently controlling the THz wave polarization which could enable applications in fast THz wave modulation, polarization-dependent THz imaging, and coherent control of nonlinear THz interaction with materials.
Fig. 8. (a) THz intensity versus THz polarizer angle and the relative phase between the and 2 pulses with linear polarization. (a) and (c) are experimental results with the two pulses parallel and orthogonally polarized, respectively; (b) and (d) are the corresponding simulation results.
Broadband THz Wave Generation, Detection and Coherent Control Using THz Gas Photonics
11
The relative phase between the and 2 pulses can be continuously tuned by using an in-line phase compensator or changing the distance between the beta barium borate crystal and laser beam focus. The intensity of the THz pulse emitted from the two-color induced plasma is measured by a pyroelectric detector. Fig. 8(a) and (c) show the change in the measured THz intensity versus THz polarizer angle and relative phase between two pulses with the polarization of the and 2 pulses parallel and orthogonal, respectively. Fig. 8(b) and (d) are the simulation results corresponding to the experiment in (a) and (c). A full quantum model is developed for the simulation. When the polarizations of two pulses are parallel, the second harmonic can lead to constructive interference on one side and destructive interference on the other side, which results in strong asymmetric electron velocity distribution. In the case of orthogonal polarized pulses, the weak second harmonic field only drives the electrons out of the oscillating direction of the fundamental beam, which leads to smaller, asymmetric electron velocity distribution compared to the parallel case. 6. Conclusion Some recent progress on THz generation, detection and coherent control in gas has been briefly covered. Due to its scalability and absence of phonon absorption or interface reflection, THz wave gas photonics has been demonstrated as a powerful tool in ultrabroadband THz spectroscopy, and as an intense and polarization-controllable THz source. With further efforts in understanding the fundamental physics and then bridging it to the latest engineering technologies, THz gas photonics should find wide application in remote THz spectroscopy, nonlinear THz spectroscopy, and in-situ diagnostics with high temporal and spatial resolution. Acknowledgement The authors gratefully acknowledge support from the National Science Foundation, Office of Naval Research, Defense Threat Reduction Agency and the Department of Homeland Security through the DHS-ALERT Center under Award No. 2008-ST-061ED0001. The views and conclusions contained in this document are those of the authors and should not be interpreted as necessarily representing the official policies, either expressed or implied, of the U.S. Department of Homeland Security.
References 1. 2. 3. 4. 5. 6. 7.
M. Tonouchi, Nat. Photonics, 1, 97-105 (2007). B. Ferguson and X.-C. Zhang, Nat. Mater. 1, 26-33 (2002). D. Mittleman, Sensing with Terahertz Radiation (Springer, Berlin, 2003). D.J. Cook and R.M. Hochstrasser, Opt. Lett. 25, 1210 (2000). M. Kress, T. Löffler, S. Eden, M. Thomson, and H.G. Roskos, Opt. Lett. 29, 1120 (2004). X. Xie, J. Dai, and X.-C. Zhang, Phys. Rev. Lett. 96, 075005 (2006). J. Dai, X. Xie, and X.-C. Zhang, Phys. Rev. Lett. 97, 103903 (2006).
12
J. Liu et al.
8. K.Y. Kim, J.H. Glownia, A.J. Taylor, and G. Rodriguez, Optics Express 15, 4577 (2007). 9. N. Karpowicz, J. Dai, X. Lu, Y. Chen, M. Yamaguchi, H. Zhao, X.-C. Zhang, L.L. Zhang, C.L. Zhang, M. Price-Gallagher, C. Fletcher, O. Mamer, A. Lesimple, and K. Johnson, Appl. Phys. Lett. 92, 011131 (2008). 10. M.D. Thomson, M. Kre , T. Loffer, and H.G. Roskos, Laser & Photon. Rev. 1, 349 (2007). 11. A. Becker, N. Akozbek, K. Vijayalakshmi, E. Oral, C.M. Bowden, and S.L. Chin, Appl. Phys. B 73, 287 (2001). 12. Q. Wu, M. Litz, and X.-C. Zhang, Appl. Phys. Lett. 68, 2924 (1996). 13. D.P. Shelton, Phys. Rev. A 42, 2578 (1990). 14. J. Dai, J. Liu, and X.-C. Zhang, IEEE J. Sel. Topics Quantum Electron. (invited) (2010). 15. J. Liu and X.-C. Zhang, Phys. Rev. Lett. 103, 235002 (2009). 16. J. Liu and X.-C. Zhang, Appl. Phys. Lett. 96, 041505 (2010). 17. J. Dai, N. Karpowicz, and X.-C. Zhang, Phys. Rev. Lett. 103, 023001 (2009). 18. H. Wen and A.M. Lindenberg, Phys. Rev. Lett. 103, 023902 (2009).
HOW DO WE LOSE EXCITATION IN THE GREEN? CHRISTIAN WETZEL*, YONG XIA, WEI ZHAO, YUFENG LI, MINGWEI ZHU, SHI YOU, LIANG ZHAO, WENTING HOU, CHRISTOPH STARK, MICHAEL DIBICCARI Future Chips Constellation and Department of Physics, Applied Physics and Astronomy, Rensselaer Polytechnic Institute, Troy, New York, U.S.A. KAI LIU, MICHAEL S. SHUR Department of Physics, Applied Physics and Astronomy, Rensselaer Polytechnic Institute, Troy, New York, U.S.A. GREGORY A. GARRETT, MICHAEL WRABACK Sensors and Electron Devices Directorate, U.S. Army Research Laboratory, Adelphi, Maryland 20783, U.S.A. THEERADETCH DETCHPROHM Future Chips Constellation and Department of Physics, Applied Physics and Astronomy, Rensselaer Polytechnic Institute, Troy, New York, U.S.A. Efficiency droop and green gap are terms that summarize performance limitations in GaInN/GaN high brightness light emitting diodes (LEDs). Here we summarize progress in the development of green LEDs and report on time resolved luminescence data of polar c-plane and non-polar m-plane material. We find that by rigorous reduction of structural defects in homoepitaxy on bulk GaN and V-defect suppression, higher efficiency at longer wavelengths becomes possible. We observe that the presence of donor acceptor pair recombination within the active region correlates with lower device performance. To evaluate the aspects of piezoelectric polarization we compare LED structures grown along polar and non-polar crystallographic axes. In contrast to the polar material we find single exponential luminescence decay and emission wavelengths that remain stable irrespective of the excitation density. Those findings render high prospects for overcoming green gap and droop in non-polar homoepitaxial growth. Keywords: light emitting diode; GaInN/GaN; green spectrum; efficiency; homoepitaxy; luminescence; time resolved spectroscopy; cathodoluminescence; transmission electron microscopy.
1. Introduction Green gap has become a term to describe the challenge to achieve visible green light by any efficient means. Replacement of energy wasteful incandescent light bulbs with solidstate lighting in the form of light emitting diodes (LEDs) has lead to a rapid progress in light generation efficiency and total light output power, yet, this has primarily been limited to the AlGaInN blue1 and the AlGaInP red2 spectral regions. In between, despite *
Electronic mail: [email protected] 13
14
C. Wetzel et al.
steady progress, the green is still trailing its spectral neighbors. For example, in 430 nm AlGaInN blue, external wall plug efficiencies up to 60% have been achieved, while 525 nm AlGaInN green hardly reaches up to 27%.3,4 This challenge in the green is compounded by the fact of an efficiency droop, common to all AlGaInN LEDs, where the light output efficiency drops substantially as the current density in the devices increases beyond some 10 A/cm2. In typical 1 W-LEDs operate at 35 A/cm2 (350 mA in 1 mm2 dies), efficiency may be only half of its maximum value. The core of our work – summarized here – aims at identifying the mechanisms that lead to either or both limiting effects and at methods to overcome those universal challenges. 2. Material and Methods As typical for high performance group-III nitride AlGaInN LEDs, our structures have been grown in metal organic vapor phase epitaxy (MOVPE), primarily along the c-axis of the uniaxial crystal structure on c-plane sapphire.5 Epitaxial growth on this mismatched substrate is initiated by low-temperature deposited nucleation layers of either AlN or GaN. Over the course of some 4 µm GaN, n-doped by Si (n = ~2x1018cm-3) for the most part, defect densities are typically reduced to the 109 cm-2 in threading dislocations (Figure 1). The active region ensues next comprising 5 to 10 periods of typically 3 nmwide Ga1-xInxN/GaN quantum wells (QWs) embedded in nominally undoped 10 to 24 nm wide GaN barriers. InN fractions x were determined by combining Θ/2ω /2θ x-ray diffraction scans of the superlattice structure with layer thickness information as obtained in high resolution transmission electron microscopy (TEM). For multiple QW samples, growth was terminated after the active region, while for full LED structures, an AlyGa1-yN (y = ~0.07) electron blocking layer p-doped ( p = ~ 1017cm-3) by Mg6 at a thickness of ~15 nm and a p-GaN layer of ~180 nm ( p = ~ 5x1017cm-3) were added.7,8
!
%&' ! " #$ " #"
%&'
$
! " #$ " #"
Fig. 1. Schematics of the different samples types employed.
How do We Lose Excitation in the Green?
15
For homoepitaxy on bulk GaN, hydride vapor phase epitaxy (HVPE) grown GaN was cut and mechano-chemically polished to epiready surfaces by collaboration partner Kyma Technologies, Raleigh, NC.9 By virtue of the high growth rate, layers sufficiently thick to be sliced along various crystallographic planes have been obtained. This unique approach allows us to compare LED performance on the polar c-plane, with those on the non-polar m- and a-planes. After HVPE growth of several hundred micrometer, the threading dislocation density along the c-growth axis can reliably be reduced into the mid 106 cm-2.9 Here we present results of materials and device performance characterization in time resolved photoluminescence (PL) spectroscopy and depth-resolved low-temperature cathodoluminescence spectroscopy. 3. Experimental Over the course of a number of development stages in our work, avoidance of spatial GaInN alloy fluctuations and any structural defects has repeatedly proven relevant to performance improvements in particular in green GaInN/GaN QW LEDs.10 As characterized in atomic force microscopy, scanning electron microscopy, and low temperature cathodoluminescence mapping, growth on low-dislocation density bulk GaN can boost 420 nm LED light output power tenfold over that of typical sapphire-based growth with dislocation densities in the mid 109 cm-2.11 Of particular concern proved to be the interaction of threading dislocations with the In-containing QWs forming V-defects that with further growth open into hexagonal pits with higher index vicinal growth surfaces. In the vicinity of such growth surface disruptions, In-incorporation and growth rate of well and barriers are found to be disturbed and, in the case of large V-defects (200 nm across), leading to very thin secondary QWs on the vicinal surfaces.12 This is accompanied by progressively wider QWs and barriers on the remaining c-plane surfaces nearby. This can be understood as a result of growth rate differences of the planes involved and an effective mass transport from the defects to the c-plane surfaces. Hangleiter et al. described such formation in short wavelength blue QWs and concluded that the higher energy QWs could provide a carrier isolation from non-radiative recombination in the threading dislocation core of the V-defect.13 For the green spectral region in particular, however, we find, that this non-uniformity leads to a rapid decay of crystalline perfection and homogeneity of the later grown QWs and, along with it, a strong decay of light output performance.14 Not until the generation of such V-defects was effectively suppressed by proper optimization of the growth processes to promote a preferential planar growth on the c-plane over vicinal plane growth, the decoration of threading dislocations with V-defect could essentially be suppressed in blue, cyan, green, and even yellow LEDs.15
16
C. Wetzel et al.
Partial Output Power (mW)
By consequential execution of the approach, light output power in respective LEDs throughout the green gap could essentially be doubled and tripled (see Figure 2).
8 2
j = 12.7 A/cm 6
With V-defects No V-defects
4 2 0 500
525
550
575
600
Dominant Wavelength (nm) Fig. 2. Achieved light output power versus dominant wavelength of the center portion of green GaInN/GaN LED wafers of individual epi runs as measured through the substrate. Wafers without V-defect formation exhibit consistently substantial higher light output performance.
Evidence, particularly in the development of green GaInN/GaN QW LEDs has repeatedly shown that photoluminescence (PL) intensity is only a weak predictor for electroluminescence (EL) light output power performance. Nevertheless, time-resolved PL decay can well serve to assess the dominant path of recombination after high density excitation. In a sequence of MQW samples at room temperature, we observe the following correlations (Figure 3): As frequently observed in such structures, the luminescence decay cannot be explained by a single decay rate, but rather follows a stretched exponential decay behavior.16 With increasing emission wavelength, the luminescence decay times increase rapidly from the ps range in the blue to the ns range in the green. Furthermore, the overall PL intensity directly scales with the luminescence decay time in the range of 30 ps to 1.4 ns. Both, the trend and the timescale of the decay time are well within expectation for a luminescence decay that is dominated by nonradiative recombination.17 The possibility to establish QW growth along both the polar and non-polar axes of GaN provides insights into the many aspects controlled by the piezoelectric polarization. In particular, since piezoelectric polarization should induce a strong electric field within the c-axis QW and change quantization and selection rules. This could reasonably play a crucial role in efficiency droop and green gap.
How do We Lose Excitation in the Green?
17
1
1
1.4ns
Decay Time (ns)
rel. PL Intensity (1)
#209 515 nm
0.1 #503 500 nm
0.01
400ps #481 480 nm
1E-3
30ps
a)
0.1
90ps Sample PeakWavelength DecayTime
p021508a.o
1E-4 0.01
#550 400 nm
0.1
1
p021508a
10
0.1
b)
Delay Time (ns)
1
PL Intensity (arb. units)
Fig. 3. a) Time resolved photoluminescence of polar c-plane GaInN/GaN MQW structures of different peak wavelength. b) The decay times scales well with the PL intensity indicating a dominance of non-radiative recombination.
We find that under certain growth regimes, non-polar m-plane and polar c-plane QWs can be grown simultaneously with similar growth rate. Figure 4 shows the results of two such grown QW structures on m-plane and c-plane GaN, respectively.18 While the polar structure shows a 558 nm deep green emission in cathodoluminescence, the nonpolar sister sample remains a 488 nm royal blue. A common interpretation for non-polar structures exhibiting such unexpected short emission wavelength is a reduced Inincorporation along such growth orientation.19,20 A detailed x-ray diffraction analysis of both lattice constants in our material, however, reveals that there is no relevant difference in InN fraction in both samples that could account for such a large 320 meV discrepancy.
MQW LED T = 300 K 30
558 nm Polar LED on c-GaN Non-polar LED on m-GaN
20
488 nm 10
0
Polarization Dipole
p05 25 07 b.o Q1 028 4A1 &A2
CL Intensity (arb. u.)
40
400
500
600
Wavelength (nm) Fig. 4. Luminescence of MQW LED structures grown simultaneously along the polar c-axis and non-polar m-axis of GaN. The discrepancy in emission energy is attributed to the polarization dipole in the polar sample (after Ref. 18).
C. Wetzel et al.
PL Intensity (cts /ms)
18
m-axis MQW LED
c-axis MQW LED Exc. density (mJ/cm 2 )
1000
517 – 502 nm
2.9 2.3 1.5
100 0.3
2.9 1.54
488 nm
10 400
a)
Excitation 337 nm Edge em ission T = 300 K
0.03
500
600
0.91
400
560 nm
2.9
500
600
Wavelength (nm) b)
Fig. 5. PL of a) non-polar m-plane and b) polar c-plane GaInN/GaN MQW structure under variable high fluence optical excitation. While emission of the polar structure shows a strong blue shift, emission of the non-polar one stays put at 488 nm (after Ref. 18).
Both structures are further analyzed under variable high density pulsed 337 nm laser excitation (Figure 5). Under conditions that can lead up to stimulated in 405 nm polar c-plane structures, the 558 nm sample shows a transition to a dominant short wavelength emission at 502 nm.18 This blue shift is commonly attributed to the screening of the quantum confined Stark effect in the piezoelectric c-plane QW. In contrast, in the non-polar structure, the emission wavelength stays pegged at 488 nm, independently of the excitation power. The advantage of longer wavelength emission in polar c-plane material obviously is limited to the low excitation density regime. Under high excitation, both, the polar and non-polar structures behave very similar. This behavior is in line with our earlier findings that the polarization dipole defined by the product of the net polarization charges at the QW interfaces with the QW width helps to reach the longer emission wavelengths in the polar structures. In absence of the piezoelectric polarization in the non-polar geometry, the dipole does not extend the wavelength of the QW. The magnitude of the polarization dipole can well be extrapolated from our earlier data21,22 and should lie around 380 meV for samples of such emission wavelength. The actual observed discrepancy in the simultaneously grown samples is in rather good agreement. Homoepitaxial growth on the various prepared surface of c-axis grown HVPE GaN provides the additional benefit of virtually suppressing all propagation of treading dislocations after the change of the growth direction. We have learned to grow full QW LED structures along all three of the c-, a-, and m-axes.7,8,11,18 For emission wavelengths up to 485 nm in m-plane, 538 nm in a-plane but below 525 nm in c-plane, we can achieve this in virtual absence of threading dislocations as judged from cross-sectional transmission electron microscopy over wide sample ranges (a quantitative assessment of density < 108 cm-2 is not possible in this method). Some dislocation lines remain at a low density below 105 cm-1.
How do We Lose Excitation in the Green? 107
0.46 5.0 42
42
104
5.0
103 102 101
a)
0.46 Fluence ( µJ/cm2)
#52 450
500
10
Fluence ( µJ/cm2)
b)
470
5
#52 0
550
Wavelength (nm)
8
PL Counts extrapol. t = 0 ns
PL Counts
105
PL Counts / Pump Power (arb. u.)
470 106
8
PLt=0 ~ Excn
10
10
7
7
10
10 n= 1
6
10
500
#50 / 0.94 #51 / 0.96 #52 / 0.95
5
10
550
0
10
Wavelength (nm)
6
10
Sample / n
c)
5
#53 / 0.85 – 1.13
2009-05-16_1827
450
19
1
2
10
10
10
3
10
Excitation power density (kW/cm2 )
Fig. 6. PL of non-polar m-plane GaInN/GaN MQW structure under variable high fluence optical excitation: a) spectra show a peak wavelength that does not shift with excitation density; b) spectra normalized to the excitation fluence show little spectral variation; c) integrated PL intensity versus excitation density.
High excitation density pulsed PL in such m-plane MQW samples is shown in Figure 6. As the excitation fluence is increased from 0.46 up to 470 µJ/cm2, the emission spectrum undergoes only minimal variation (Figure 6a). Figure 6b) shows the spectra after normalization to the excitation fluence. The integrated luminescence output versus excitation power density (Figure c), shows a very close to linear correlation. The time resolved PL decay of the same structures is shown in Figure 7. As the excitation fluence is increased over four orders of magnitude, the luminescence decay shows only minimal variation in the decay behavior (Figure 7a). The spectral variation of the decay times across the set of samples (Figure 7b) shows simple single exponential decay described by decay times varying only between 100 to 200 ps. Similar to the case of the polar structures, this behavior is possibly dominated by non-radiative recombination. 7
8
10
10
5 µJ/cm
2
#52 470 nm
7
6
470
6
10
42
10
PL Signal (cts)
PL Signal (cts)
10
5.0
5
10
0.46 4
10
Fluence (µJ/cm2)
3
10
PLmax / τdecay
5
443 nm / 227 ps 470 nm / 196 ps 475 nm / 140 ps 485 nm / 105 ps
10
4
10
3
10
2
10
2
0
a)
1
2
Time (ns)
3
10
b)
0
1
2
3
Time (ns)
Fig. 7. Time resolved luminescence decay in non-polar m-plane GaInN/GaN MQW structures: a) a single exponential decay is observed independent of excitation fluence; b) as a function of center wavelength, only a minimal variation of the decay time is observed.
20
C. Wetzel et al.
c
LED on bulk GaN IDPs 9 -2
~6x10 cm
LED on sapphire No MDs 9
515 nm, new threading dislocations are being initiated in the highly strained QWs.7,23 Figure 8 shows the light output power as a function of wavelength in a series of green and deep green LEDs. Associated with the data also are cross-sectional TEM micrographs revealing some sample-characteristic dislocation networks. The data includes samples on sapphire (red symbols) and on bulk GaN (green symbols). It can be seen that while the presence of threading dislocations has not been suppressed in structures on sapphire, the generation of V-shaped defects is relegated to the very long wavelength range as seen in a 589 nm yellow-orange structure. While reduction of defects observable in TEM has resulted in significant performance improvements, particularly in the green spectral region, it is doubtful, that dislocations are the only culprit for non-radiative recombination. This becomes particular apparent from the persisting droop performance in those non-polar samples, where threading dislocation densities have been pushed below the detection limit of 108 cm-2.7,8
CL Intensity (arb. units)
6 5 4 3
Blue QW
GaN NBE
Green QW
DAP
1
400
450
500
550
600
CL Intensity (arb. units) .
3
Sample 4 T = 77K
3kV > 1 and s / L >> 1, where and are the angular frequency of the incident radiation and the momentum relaxation time respectively, s is the plasma wave velocity and L is the device channel length. The plasma wave velocity is shown to be dependent upon the gate bias as s = (q Vgt /m)1/2, where q is the electronic charge, m is the electron effective mass and Vgt = Vgs – Vt is the gate bias swing relative to the threshold voltage Vt. Thus such resonant detectors are tunable via the gate bias. For devices in which >> 1 but s /L Neff, the resistance of the passive fins is found as:
RP =
1
(N − N ) eff
RS + RD +
( 2H
Lg
f
+ W f ) q µ ns
(21)
34
W. Stillman et al.
We now consider the effective circuit for the DC THz response as shown in Figure 3(a), and the Thevenin equivalent circuit in Figure 3(b), where:
VTh = δ v
RP RP + RA
(22)
and:
RTh =
RA RP RA + RP
(23)
Figure 3 – (a) DC equivalent circuit for FinFET, and (b) Thevenin equivalent representation.
Hence, the detector response is now given by:
δ veff =
va2 RP RL sinh 2 κ − sin 2 κ 4ms 2 sinh 2 κ + cos 2κ RL ( RA + RP ) + RP RA
(24)
Until this point, we have ignored the effects of coupling of the incident radiation to the device. While there certainly are a great many factors to consider, for our purpose of modeling the frequency response of the device without an optimized antenna, we consider only the additional effects of the bond wire inductance on the response. We model the attenuation of the incident signal using a lumped parameter voltage divider between the device capacitance and series connected device resistance and bond wire inductance, resulting in the following expression for the effective coupled radiation:
va = vTHz
Zo Z o + ZTHz
(25)
Silicon FinFETs as Detectors of Terahertz and Sub-Terahertz Radiation
35
Where Zo is given by eq. (15). We assume that the impedance of the effective THz source is inductive with the effective inductive impedance given by:
ZTHz = jω LTHz
(26)
Combining eqs. (24) and (25) our response expression becomes:
δ veff =
2 vTHz sinh 2 κ − sin 2 κ 4ms 2 sinh 2 κ + cos 2κ
RP RL RL ( RA + RP ) + RP RA
Zo Z o + ZTHz
2
(27)
It is expected that this expression will be applicable to the drain current enhanced response of eqs. (6) and (7) with the appropriate adjustments as discussed previously. The speed of detector response is also of importance. Kachorovskii and Shur [18] propose theoretical calculations of the maximum response modulation frequency in plasma wave detectors as shown in the following expressions:
µ fVgt 2π L2g
Vgt > 0, qVgt >> T
f max =
(28)
µ f η k BT Vgt < 0, q Vgt >> T 2π qL2g Here f is the effective field effect mobility, which is very different from conventional mobility for short channel devices, where ballistic or near ballistic transport is dominant. Near Vgt = 0, fmax is interpolated as: f max =
µ f η k BT 2π qL2g
1 + exp
−qVgt
η k BT
ln 1 + exp
qVgt
η k BT
(29)
4. Response Measurements Prior to response measurements, current / voltage characteristics were performed and die containing several functional devices were wire bonded within ceramic chip carriers to simplify handling. Response measurements were made using a standard lock-in technique. Two radiation source types were used: a purpose built Gunn diode oscillator equipped with frequency multipliers was used at the 0.2 and 0.6 THz. An optically pumped terahertz gas laser was used at 1.6 and 2.4 THz. Source power was measured as 1.5 mW, 35 W, 30 mW and 10 mW for 0.2, 0.6, 1.6 and 2.4 THz respectively. A chopper was placed in the beam path and the radiation was focused onto the device using either a parabolic mirror or polyethylene lens depending upon the source. Two programmable power supplies were used to provide gate and drain bias. Figure 4 illustrates the typical response dependence upon gate bias.
36
W. Stillman et al.
Figure 4 – 20 fin nFET open drain response, normalized to response maxima, across several incident frequencies. Wf = 40 nm; Lg = 100 nm. Dashed line represents modeled response using eq. (5) with f = 0.2 THz.
The response at each frequency is normalized to its peak value to allow direct comparison across the frequency range, and good agreement with the model of eq. (5) is apparent. That the response peaks coincide across the frequency range confirms the nonresonant nature of the response. Figure 5 compares responsivity modeled using eq. (27) to measured data for several FinFET devices. (Calculations of responsivity in measured data are simply measured response divided by incident power; no adjustment for device vs. beam size is used since it is not clear that this approach is warranted [19]). Note that while there are anomalies, the fit is reasonable, especially in the response attenuation at and above 0.6 THz. In addition, note that the FinFET devices in many cases exhibit considerably greater responsivity than standard CMOS FETs.
Silicon FinFETs as Detectors of Terahertz and Sub-Terahertz Radiation
37
Figure 5 – Open drain responsivity of several FinFET devices at various incident frequencies. Filled symbols are measured data; lines are modeled response using eq. (27) with Lg = 100 nm, Lpitch = 200 nm, Hf = 40 nm, Wf = 60 nm, / = 6.9x10-2 F/m2, g = 10 ohms/ , RS = RD = 300 ohms, = 0.5 m2/Vs, Vgt = -0.1 V. VTHz and LTHz were chosen to fit measured data. Open symbols are measured responsivity for standard CMOS FETs. [19, 20]
Responsivity rises dramatically as expected in FinFET devices, reaching from several hundred to above one thousand volts per watt. Figure 6 illustrates the degree to which the enhanced responsivity follows that predicted with eq. (6). In Figure 7 is shown the drain current enhanced responsivity for several FinFETs at various frequencies compared with the response of standard CMOS FET data. Here especially is demonstrated the advantage of the FinFET structure, as responsivity is seen to be nearly two orders of magnitude higher than the standard CMOS FETs.
38
W. Stillman et al.
Figure 6 – 0.2 THz responsivity of 20 fin device with Wf = 40 nm and Lg = 100 nm. Symbols are measured data; lines are modeled responsivity following eq. (6).
Figure 7 – Peak drain current enhanced responsivity of several FinFET devices vs. incident frequency. Filled symbols are measured data. Dashed line is drawn to guide the eye. Open symbols are measured responsivity for standard CMOS FETs. [20]
Silicon FinFETs as Detectors of Terahertz and Sub-Terahertz Radiation
39
5. Noise Equivalent Power Noise Equivalent Power represents the minimum signal power distinguishable from the detector noise, and is a figure of merit for the detector sensitivity. Commonly, NEP is considered to be the minimum power detectable per square root of bandwidth, in units of W/ Hz, and is calculated as the inverse ratio of responsivity to the square root of device voltage noise. In the open drain configuration, the thermal noise of the channel resistance is predominant [10, 21]. Reducing the channel resistance decreases NEP for a given responsivity by a factor of the square root of the resistance reduction, thus increasing the number of device fins from 2 to 20, or 20 to 200 will decrease the noise contribution to NEP by a factor of ~3.2. Returning to Figure 5 however, finds a decrease in responsivity of one magnitude order between 20 and 200 fin devices, therefore NEP increases for the larger number of fins. This is illustrated in Figure 8. The comparison between 2 and 20 fins is less clear due to the observed anomalies in responsivity for the 2 fin devices, though in general NEP is expected again to be higher for the larger number of fins.
Figure 8 – Noise equivalent power vs. gate bias for two FinFET devices at 0.2 (filled symbols) and 0.6 THz (open symbols) calculated from measured response and channel resistance data. Wf = 40 nm, Lg = 60 nm, 20 fins (triangles) and 200 fins (squares).
Estimation of NEP becomes more complicated with the introduction of drain current. Responsivity increases on average between 20 and 30 times that of the open drain response for the FinFET devices; however device noise increases dramatically as well, essentially as a function of the drain current squared. The noise spectra are seen to follow
40
W. Stillman et al.
a 1/f a distribution, with a 1, as is shown in Figure 9. Here the spectrum measured with a drain load resistance of 2.5 kohms is used to estimate the spectra with loads more typical of those used in responsivity measurements, allowing comparison of NEP with drain current to that of the open drain configuration. Peak responsivity for the device in Figure 9 was measured to be 1200 V/W; at the optical chopper frequency used in response measurements of 50 Hz and with an amplifier load of approximately 10 Mohm, the voltage noise density is calculated to be 5 x 10-8 V2/Hz, thus the NEP at 50 Hz is approximately 2 x 10-7 W/Hz1/2. The sampling frequency required to achieve the open drain NEP for this device of 3 x 10-3 W/Hz1/2 is calculated to be approximately 35 MHz, well below the ~30 GHz maximum response frequency predicted by eq. (29).
Figure 9 – Voltage noise spectral density for 20 fin device with Wf = 40 nm and Lg = 100 nm. Measured data at Vgt = 0V and Id = 3.1 A with 2.5 kohm drain load (lowest line) is calculated to show expected device noise for larger values of load resistance (responsivity measurements presented earlier are loaded at 10 Mohm). Horizontal dashed lines are corresponding calculated thermal noise values; angled dashed lines project device noise to these values to indicate frequencies of equivalence.
6. Conclusion We have demonstrated the response of Si FinFETs to terahertz and sub-terahertz radiation and developed a model to explain the observed decrease in response as the number of device fins increases, which we attribute to the attenuation of the coupled incident radiation both across the device gate conductor transmission line and due to the inductance of the device bond wires. Our results show that narrow FinFETs can be
Silicon FinFETs as Detectors of Terahertz and Sub-Terahertz Radiation
41
competitive or better than commercial THz detectors potentially enabling the development of sub-THz and THz cameras implemented using this technology.
Acknowledgments The work at RPI is supported by SEMATECH, by the InterconnectFocusCenter, and by the NSF under the auspices of the I/UCRC “CONNECTION ONE”.
References [1] W. J. Stillman and M. S. Shur, “Closing the Gap: Plasma Wave Electronic Terahertz Detectors,” Journal of Nanoelectronics and Optoelectronics, vol. 2, pp. 209-221, 2007. [2] T. A. Elkhatib, A. V. Muravjov, D. B. Veksler, W. J. Stillman, X.-C. Zhang, M. S. Shur, and V. Y. Kachorovskii, “Subwavelength Detection of Terahertz Radiation using GaAs HEMTs,” Proceedings of IEEE Sensors, pp. 1988-1990, 2009. [3] QMC_Instruments, http://www.terahertz.co.uk/,2010 [4] Spectrum_Detector_Inc., http://www.spectrumdetector.com/pdf/datasheets/THZ.pdf,2007 [5] Virginia_Diodes_Inc,http://virginiadiodes.com/WR2.2ZBD.htm,2007 [6] M. Dyakonov and M. Shur, “Shallow water analogy for a ballistic field effect transistor: New mechanism of plasma wave generation by dc current,” Physical Review Letters, vol. 71, pp. 2465-2468, 1993. [7] M. I. Dyakonov and M. S. Shur, “Plasma wave electronics: Novel terahertz devices using two dimensional electron fluid,” IEEE Transactions on Electron Devices, vol. 43, pp. 1640-1645, 1996. [8] M. Dyakonov and M. Shur, “Detection, mixing, and frequency multiplication of terahertz radiation by two-dimensional electronic fluid,” IEEE Transactions on Electron Devices, vol. 43, pp. 380-387, 1996. [9] W. Knap, V. Kachorovskii, Y. Deng, S. Rumyantsev, J. Q. Lu, R. Gaska, M. S. Shur, G. Simin, X. Hu, M. Asif Khan, C. A. Saylor, and L. C. Brunel, “Nonresonant detection of terahertz radiation in field effect transistors,” Journal of Applied Physics, vol. 91, pp. 9346, 2002. [10] W. Stillman, M. S. Shur, D. Veksler, S. Rumyantsev, and F. Guarin, “Device loading effects on nonresonant detection of terahertz radiation by silicon MOSFETs,” Electronics Letters, vol. 43, pp. 422-423, 2007. [11] J. Q. Lu and M. S. Shur, “Terahertz detection by high-electron-mobility transistor: Enhancement by drain bias,” Applied Physics Letters, vol. 78, pp. 2587, 2001. [12] D. Veksler, F. Teppe, A. P. Dmitriev, V. Y. Kachorovskii, W. Knap, and M. S. Shur, “Detection of terahertz radiation in gated two-dimensional structures governed by dc current,” Physical Review B (Condensed Matter and Materials Physics), vol. 73, pp. 1253281, 2006. [13] M. Schubert, B. Hofflinger, and R. P. Zingg, “A one-dimensional analytical model for the dual-gate-controlled thin-film SOI MOSFET,” Electron device letters, vol. 12, pp. 489-491, 1991. [14] D. J. Frank, S. E. Laux, and M. V. Fischetti, “Monte Carlo simulation of a 30 nm dual-gate MOSFET: how short can Si go?,” International Electron Devices Meeting 1992. Technical Digest (Cat. No.92CH3211-0), pp. 553-6, 1992. [15] K. Suzuki, T. Tanaka, Y. Tosaka, H. Horie, and Y. Arimoto, “Scaling theory for double-gate SOI MOSFET’s,” IEEE Transactions on Electron Devices, vol. 40, pp. 2326-9, 1993.
42
W. Stillman et al.
[16] C. Fiegna, H. Iwai, T. Wada, M. Saito, E. Sangiorgi, and B. Ricco, “Scaling the MOS transistor below 0.1 m: methodology, device structures, and technology requirements,” IEEE Transactions on Electron Devices, vol. 41, pp. 941-51, 1994. [17] D. Hisamoto, L. Wen-Chin, J. Kedzierski, H. Takeuchi, K. Asano, C. Kuo, E. Anderson, K. Tsu-Jae, J. Bokor, and H. Chenming, “FinFET-a self-aligned double-gate MOSFET scalable to 20 nm,” IEEE Transactions on Electron Devices, vol. 47, pp. 2320-5, 2000. [18] V. Y. Kachorovskii and M. S. Shur, “Field effect transistor as ultrafast detector of modulated terahertz radiation,” Solid State Electronics, vol. 52, pp. 182-5, 2008. [19] W. J. Stillman, “Silicon CMOS FETs as terahertz and sub-terahertz detectors,” in Doctoral Thesis. Troy, NY: RPI, 2008. [20] W. Stillman, F. Guarin, V. Y. Kachorovskii, N. Pala, S. Rumyantsev, M. S. Shur, and D. Veksler, “Nanometer scale complementary silicon MOSFETs as detectors of terahertz and sub-terahertz radiation,” IEEE Sensors 2007 Conference, pp. 934-7, 2007. [21] R. Tauk, F. Teppe, S. Boubanga, D. Coquillat, W. Knap, Y. M. Meziani, C. Gallon, F. Boeuf, T. Skotnicki, C. Fenouillet-Beranger, D. K. Maude, S. Rumyantsev, and M. S. Shur, “Plasma wave detection of terahertz radiation by silicon field effects transistors: Responsivity and noise equivalent power,” Applied Physics Letters, vol. 89, pp. 253511, 2006.
PROGRESS IN DEVELOPMENT OF ROOM TEMPERATURE CW GASB BASED DIODE LASERS FOR 2-3.5 µM SPECTRAL REGION TAKASHI HOSODA Department of Electrical and Computer Engineering, State University of New York at Stony Brook, NY 11794, USA [email protected] JIANFENG CHEN Department of Electrical and Computer Engineering, State University of New York at Stony Brook, NY 11794, USA [email protected] GENE TSVID Department of Electrical and Computer Engineering, State University of New York at Stony Brook, NY 11794, USA [email protected] DAVID WESTERFELD Power Photonic Corporation, Stony Brook, New York 11790, USA [email protected] RUI LIANG Department of Electrical and Computer Engineering, State University of New York at Stony Brook, NY 11794, USA [email protected] GELA KIPSHIDZE Department of Electrical and Computer Engineering, State University of New York at Stony Brook, NY 11794, USA, Power Photonic Corporation, Stony Brook, New York 11790, USA [email protected] LEON SHTERENGAS Department of Electrical and Computer Engineering, State University of New York at Stony Brook, NY 11794, USA [email protected] GREGORY BELENKY Department of Electrical and Computer Engineering, State University of New York at Stony Brook, NY 11794, USA, Power Photonic Corporation, Stony Brook, New York 11790, USA [email protected]
43
44
T. Hosoda et al. Recent progress and state of GaSb based type-I lasers emitting in spectral range from 2 to 3.5 µm is reviewed. For lasers emitting near 2 µm an optimization of waveguide core width and asymmetry allowed reduction of far field divergence angle down to 40-50 degrees which is important for improving coupling efficiency to optical fiber. As emission wavelength increases laser characteristics degrade due to insufficient hole confinement, increased Auger recombination and deteriorated transport through the waveguide layer. While Auger recombination is thought to be an ultimate limiting factor to the performance of these narrow bandgap interband lasers we demonstrate that continuous improvements in laser characteristics are still possible by increasing hole confinement and optimizing transport properties of the waveguide layer. We achieved 190, 170 and 50 mW of maximum CW power at 3.1, 3.2 and 3.32 µm wavelengths respectively. These are the highest CW powers reported to date in this spectral range and constitute 2.5-fold improvement compared to previously reported devices. Keywords: GaSb; InGaAsSb; type-I; diode lasers; quantum well lasers; midinfrared; long wavelength.
1. Introduction Semiconductor lasers emitting in 2 to 3.5 µm wavelength range are in demand for laser spectroscopy, medical diagnostics, medical therapy and solid state laser pumping [1-5]. This demand is being met by developing GaSb based type-I diode laser technology. At the short wavelength end of the region InGaSb quantum well lasers on GaSb substrate show excellent performance. High power 2 m diode lasers with broadened waveguide achieved threshold current densities of 100 A/cm2. Differential quantum efficiencies of 53% for 2-mm-long multimode devices were demonstrated [6-7]. Unfortunately broadened waveguide lasers have rather large fast axis divergence, typically about 65 degrees full width at half maximum (FWHM). To facilitate coupling of light to an optical fiber a divergence of 40 to 50 degrees is desirable. To this end we present an optimization of waveguide core width and asymmetry for 2 µm emitting devices. By adjusting waveguide and cladding parameters the fast axis divergence of 45-50 degrees and 1 W output power with 15% power conversion efficiency can be achieved. In the second part of the paper we review CW power performance over spectral range of 2 to 3.4 µm. Until recently the prospect of making diode lasers operating in CW at RT with wavelength near 3 m was thought to be questionable. This pessimistic scenario originated from recognition of the well known fundamental increase of the probability of the nonradiative Auger recombination and free carrier absorption with wavelength. The associated carrier and photon losses were considered to prevent the mid-infrared diode lasers from reaching CW threshold at RT. However we demonstrate that optimization of waveguide composition and thickness results in improved performance. In particular devices with 32% of indium in 640 nm wide quinary InAlGaAsSb waveguide produced maximum CW powers of 190, 170 and 50 mW at 3.1, 3.2 and 3.32 µm wavelengths respectively. 2. Results and Discussion 2.1. Waveguide core width and asymmetry optimization for 2 µm emitting devices The laser heterostructures were grown by solid-source molecular beam epitaxy in a Veeco GEN-930 reactor equipped with As and Sb valved cracker sources on GaSb
Development of Room Temperature CW GaSb Based Diode Lasers
45
Te-doped substrates. Te and Be were used for n and p doping respectively. The laser active region comprised two 12 nm wide In0.2Ga0.8Sb quantum wells separated by 20 nm of AlGaAsSb quaternary alloy lattice matched to GaSb substrate. The devices were nominally identical except for variations in waveguide core and cladding widths and aluminum content of n-cladding as shown in Table 1.
Table 1. Aluminum composition and width of the claddings and core layers of 2 m emitting lasers. Sample ID % Al 1 2 3 4 5 6
90 90 90 90 90 90
p-cladding Thickness [nm] 1500 1500 1500 1500 1500 1000
% Al
Core Thickness [nm]
% Al
25 25 25 25 25 30
859 850 550 350 250 450
90 50 50 50 50 50
n-cladding Thickness [nm] 1500 2000 2000 2000 2500 2500
Two parameters varied in this series are the waveguide core width and asymmetry of the mode field relative to N and P claddings. As expected the narrower waveguide core the larger is the mode field waist and hence smaller far field divergence angle, Figure 1f. The penalty for larger penetration of the mode field into the claddings is an increase in internal losses. Since free hole absorption is expected to be higher than that of free electrons [8-9] it is beneficial to design asymmetric mode that penetrates into the n-cladding more than into the p-cladding. This can be done by lowering aluminum content of the n-cladding layer, increasing n-cladding thickness and decreasing p-cladding thickness. Continuous wave power characteristics of selective structures are given in Figure 1. Diode lasers with structure 3 (far field divergence with FWHM of 50 degrees) generate above 1.5 W of CW power and demonstrate peak power conversion efficiency in excess of 28%, Figure 1a. Linear laser array with fill factor of 20% composed of eighteen 2-mm-long emitters generated above 11 W at 20C at current of 55 A, Figure 1b. Power conversion efficiency was above 20% in peak and above 15% at maximum power level. Driving current and output power were limited by TEC stage heat removal capabilities. Diode lasers with structure 4 (far field divergence with FWHM of 44 degrees) generate 1.4 W of CW power but peak power conversion efficiency was below 20%, Figure 1c. Smaller value of the output power and efficiency is associated with higher optical loss and stronger temperature sensitivity of the threshold. Still the devices were producing more than watt of 2 µm power at power conversion efficiency better than 10%. Seven 2-mm-long AR/HR coated single emitters of structure 4 were epi-down mounted onto Au-coated BeO blocks and connected in series.
30
1.5 20 1.0 10 0.5
0.0
0
2
2.00 λ (µm)
4 Current (A)
2.05
6
8
0
15 1.0 10 0.5 5
0.0
0
2
2.00 λ (µm)
4 Current (A)
2.05
6
8
0
(c) CW power from 100 µm edge emitter, structure 4. 30
1-mm-long 2-mm-long
1.0
20
0.5
10 1.95
0.0
0
2
2.00
2.05
λ (µm)
4 Current (A)
10
5 Near Field Pattern 0
0
20
5 0 60
40
6
8
0
(b) CW power from 18 element array, structure 3.
3
2
Near Field Pattern
1
0
0
2
4 Current (A)
6
8
(d) CW power from 7 element fiber bundle, structure 4. 1.0
Power Conversion Efficiency (%)
Output Power (W)
1.5
15
4
Power Conversion Efficiency (%)
Output Power (W)
20
1-mm-long 2-mm-long
1.95
20
Current (A)
(a) CW power from 100 µm edge emitter, structure 3. 1.5
25
18*2-mm-long emitters 20% fill factor
10
Output Power (W)
1.95
15
(e) CW power from 100 µm edge emitter, structure 6.
Intensity (a.u.)
Output Power (W)
1-mm-long 2-mm-long
Output Power (W)
2.0
Power Conversion Efficiency (%)
T. Hosoda et al. Power Conversion Efficiency (%)
46
0.5
0.0 -90
str 1 str 2 str 3 str 4 str 5
-60
-30 0 30 60 Fast Axis Angle (Degree)
90
(f) Far field fast axis divergence fields of structures 1 to 5.
Figure 1. Power outputs and power conversion efficiencies subfigures (a,c,e) of structures (3,4,6) respectively; array CW power outputs of structures 3,4, subfigures (b,d); fast axis far fields for structures 1 to 5 (f).
Development of Room Temperature CW GaSb Based Diode Lasers
47
Output light was collected by a cylindrical micro-lens in front of each emitter into seven multimode silica fibers (core diameter of 105 m). No optimization of the collecting optics was done for the particular 2 µm laser technology. More than 3 W of CW power was collected from the other end of the fiber bundle, Figure 1d. For structure 6 QWs were shifted closer to the n cladding layer. As low as 100 A/cm2 and 85 A/cm2 of CW threshold current density, at 20C, were achieved for 1 and 2-mm-long AR/HR coated lasers, respectively, Figure 1e. Output power above 1.3 W was generated at 8 A. Power conversion efficiency peaked above 25% but rapidly decreased with current due to voltage drop across series resistance. Increase of the series resistance in structure 6 as compared to structure 3, for instance, is not fully understood. The excessive voltage drop across laser heterostructure can be ascribed to reduced doping in p-cladding layer (1017 cm-3 of Be was used in p-cladding of structure 6 as compared to 2*1017 cm-3 in structures 1-4). This model would have to assume rather low hole mobility. Another possible explanation is an adverse variation of the substrate doping level and associated increase of the n-contact resistance. 2.2. CW power characteristics over 2.2 to 3.3 µm wavelength range By increasing indium composition in the quantum well and adding arsenic as necessary to keep strain within permissible 2% level it is possible to extend emission wavelength up to 3.4 micron. Figure 2 shows CW power characteristics achievable at 2.2, 2.7, 3.0, 3.1, 3.2 and 3.32 µm emission wavelengths. Maximum CW powers are approximately 1500, 600, 300, 190, 170 and 50 mW. This degradation is due to an increase in Auger recombination rate [10], decrease in hole confinement [11] and, as more recently discussed, carrier recombination in the waveguide [12]. At 2.2 µm CW power of 1 Watt can be achieved with 17% wall-plug efficiency while at 2.7 µm we get 400 mW with 9% of power conversion efficiency. Recently it was found that two improvements can be made to devices emitting above 3 µm. First, due to carrier recombination in the waveguide narrower cores are preferred [12]. Second, hole confinement can be improved by increasing indium composition in the waveguide [13]. Application of these two concepts to lasers emitting above 3 µm resulted in CW characteristics shown in Figure 2b, [14]. Devices differ in indium composition in the quantum wells to achieve emission at 3.0, 3.1, 3.2 (3.2A and 3.2B) and 3.32 µm. Lasers 3.2A and 3.2B, both emitting near 3.2 µm, had indium concentrations of 25% and 32% in quinary AlInGaAsSb waveguide layer respectively. The laser 3.2B with 32% of indium in the waveguide produced 1.5 times more maximum CW power than the laser with 25% of indium, 3.2A, as we believe, due to increased hole confinement.
T. Hosoda et al.
15 1.0 10 0.5 2.15
0
2
2.20
λ (µm)
4
5
2.25
6
8
0
12 10
0.6 8 6
0.4
4 0.2 2.6 0.0
Current (A)
0
2
2.7
4
2
2.8
λ (µm)
6
8
Power Conversion Efficiency (%)
20
1.5
0.0
0.8
25
Output Power (W)
Output Power (W)
2.0
Power Conversion Efficiency (%)
48
0
Current (A)
A) CW power (left axis) and power conversion efficiency (right axis) for lasers emitting at 2.2 and 2.7 µm. Insets are emission spectra.
Output Power (W)
0.4
3.0 3.1 3.2B 3.2A 3.3
3.0
0.3 2.9
3.0
3.1 3.2 λ (µm)
3.3
3.4
0.2
B) CW powers of 3.0, 3.1, 3.2 (3.2AAl0.22In0.25Ga0.53AsSb and 3.2B- Al0.22In0.32Ga0.46AsSb lattice matched to GaSb waveguide layer) and 3.32 m emitting 2-mm-long AR/HR coated lasers. The inset shows the laser spectra.
3.1 3.2B
0.1
Figure 2. CW power performance of lasers in spectral region 2.2 - 3.3 µm. Inset shows emission spectra.
3.2A 3.3
0.0
0
1
2 3 Current (A)
4
5
3. Conclusion We reviewed design strategies and characteristics of the best InGaAsSb quantum well lasers emitting in 2 to 3.3 µm wavelength range. For 2 µm emitting lasers fast axis far field divergence angles of 45 to 50 degrees are attainable with only minor sacrifice to CW power and power conversion efficiency. Corresponding single emitters with 100 µm wide output aperture generate above 1 W at 20 C of 2 µm optical power with power conversion efficiency exceeding 15% at maximum power level and peaking at 28%. Linear laser arrays generate above 10 W of CW power. More than 3 W of CW power was demonstrated from fiber bundle containing seven 105- m diameter silica fibers. Significant improvement to devices emitting above 3 µm was demonstrated. In particular lasers with 640 nm thick, 32% indium, quinary AlInGaAsSb waveguide layer have produced maximum CW powers of 300, 190, 170 and 50 mW for 3.0, 3.1 3.2 and 3.32 µm emission respectively. The improvement is attributed to increase in hole confinement and optimal width of the waveguide.
Development of Room Temperature CW GaSb Based Diode Lasers
49
Acknowledgements The work was supported by Air Force Office of Scientific Research by Grant FA95500810458 and Young Investigator Award FA95500810083, National Science Foundation Grant DMR0710154 and US Army Research Office under Contract W911NF0610399. References 1. Curl, R.F., et al., Quantum cascade lasers in chemical physics. Chemical Physics Letters, 2010. 487(1-3): p. 1-18. 2. Elia, A., et al., Photoacoustic Techniques for Trace Gas Sensing Based on Semiconductor Laser Sources. Sensors, 2009. 9(12): p. 9616-9628. 3. Lewicki, R., et al., Carbon dioxide and ammonia detection using 2 m diode laser based quartz-enhanced photoacoustic spectroscopy. Applied Physics B: Lasers and Optics, 2007. 87(1): p. 157-162. 4. Mond, M., et al., 1.9-mm and 2.0-mm laser diode pumping of Cr2+: ZnSe and Cr2+: CdMnTe. Opt. Lett, 2002. 27: p. 1034-1036. 5. Sorokina, I. and K. Vodopyanov, Solid-state mid-infrared laser sources. 2003: Springer. 6. Garbuzov, D., et al., 4 W quasi-continuous-wave output power from 2 m AlGaAsSb/InGaAsSb single-quantum-well broadened waveguide laser diodes. Appl. Phys. Lett., 1997. 70: p. 2931. 7. Turner, G., H. Choi, and M. Manfra, Ultralow-threshold (50 A/cm) strained single-quantumwell GaInAsSb/AlGaAsSb lasers emitting at 2.05 m. Appl. Phys. Lett., 1998. 72: p. 876. 8. Chandola, A., R. Pino, and P. Dutta, Below bandgap optical absorption in tellurium-doped GaSb. Semicond. Sci. Technol., 2005. 20: p. 886. 9. Rattunde, M., et al., Comprehensive analysis of the internal losses in 2.0 mu m (AlGaIn)(AsSb) quantum-well diode lasers. Appl. Phys. Lett., 2004. 84(23): p. 4750-4752. 10. Haug, A., Auger recombination in direct-gap semiconductors: band-structure effects. Journal of Physics C: Solid State Physics, 1983. 16: p. 4159-4172. 11. Shterengas, L., et al., Design of high-power room-temperature continuous-wave GaSb-based type-I quantum-well lasers with lambda > 2.5 mu m. Semicond. Sci. Technol., 2004. 19(5): p. 655-658. 12. Hosoda, T., et al., 200 mW type I GaSb-based laser diodes operating at 3 µm: Role of waveguide width. Appl. Phys. Lett., 2009. 94(26). 13. Shterengas, L., et al., Diode lasers emitting at 3 µu m with 300 mW of continuous-wave output power. Electronics Letters, 2009. 45(18): p. 942-U37. 14. Hosoda, T., et al., Type-I GaSb-Based Laser Diodes Operating in 3.1- to 3.3 µm Wavelength Range. Photonics Technology Letters, IEEE, 2010. 22(10): p. 718-720.
This page intentionally left blank
WDM DEMULTIPLEXING BY USING SURFACE PLASMON POLARITONS DJAFAR K. MYNBAEV* Department of Electrical and Telecommunications Engineering Technology New York City College of Technology of the City University of New York 300 Jay Street, Brooklyn, NY 11201, USA [email protected] VITALY SUKHARENKO Department of Electrical and Telecommunications Engineering Technology (student) New York City College of Technology of the City University of New York 300 Jay Street, Brooklyn, NY 11201, USA The volume of telecommunications traffic keeps growing at an exponential rate. The opticalcommunications industry, the linchpin of modern telecommunications, in its quest of keeping up with this growth simply must increase the number of wavelengths in the wavelength-division multiplexing (WDM) configuration. The result of this increase would mean, too, that the number of transmitters and receivers that could be placed on one board would increase as well; hence, the density of their packaging would come to micro- and even nano-scale. At the receiver end, which we will consider in this paper, manufacturers are now able to fabricate an array of photodiodes (PDs) on a single wafer, reducing the size of an individual PD to hundreds of nm. The main obstacle in approaching this scale from a communications link point of view, however, is the diffraction limit. One possible solution to this problem is the use of plasmonics. This paper discusses a possible approach to using surface plasmon polaritons (SPPs) for WDM demultiplexing, presents a possible scheme to implement this approach, and offers our analysis of this scheme along with suggestions for future developments. Keywords: Surface plasmon polaritons; WDM demultiplexing.
1.
Introduction
Surface plasmon polaritons (SPPs) are two-dimensional waves that propagate between conductors (metals) and dielectrics. These surface waves are excited when light strikes the dielectric-metal interface; the energy of the photons is transferred to the metal and resonantly excites the oscillations of free electrons. The electrons’ response results in the creation of dynamic charges on the metal’s surface; these charges, in turn, produce waves called SPPs [1]. Fig. 1 illustrates these explanations. SPPs can be considered, in essence, secondary EM radiation obtained in response to incident light.
*
New York City College of Technology, 300 Jay Street, Brooklyn, NY 11201, USA. 51
52
D. K. Mynbaev & V. Sukharenko
Fig. 1. This Surface plasmon polaritons: Electron charges and excited electric waves. (Reprinted with permission of Igor Zozulenko2.)
The term surface refers to the waves that travel along the metal surface (between the metal and a dielectric) in contrast to waves excited in the metal volume. The term plasmon stems from the word plasma, a charged gas describing the behavior of free electrons according to the Drude model of metals. Polariton is a quasiparticle—half wave and half matter—well defined in physics. SPPs have been well known for more than a century; today’s strong interest in the study of this phenomenon has been due to the latest advances in nanotechnology, advances that now make applications of SPPs in numerous fields quite feasible. The main theory behind the excitation of SPPs, as well as the techniques for their excitation, is well known1; we will consider here only the excitation of SPPs with diffraction gratings. 2.
The Frequency of SPPs
We consider here a thin metal film placed between two dielectrics with corrugations on both sides of the film; thus, we have top and bottom metallic gratings. In this paper, we consider light falling on the top grating and SPP radiation emanating from the bottom grating. Fig. 2 shows an example of such an arrangement and excited SPPs3. When the metal strip is thicker than 100 nm, only a single SPP mode gets excited; when, however, the metal film is thinner than 100 nm, two SPP modes—short-range and long-range—are excited.
WDM Demultiplexing by Using Surface Plasmon Polaritons
53
Fig. 2. SPP modes excited on a thin metal strip corrugated on both sides (metal gratings): (a) and (b) show the actual pictures of the excited SPPs; (c) and (d) schematically show charges and excited SPP modes. SRSPP and LRSPP stand for short-range SPP and long-range SPP, respectively. (Reprinted with permission from3.)
Thus, in general, a thin metal strip with gratings might acts as a transducer of incident photons to SPPs. The goal of this section of our paper is to present the frequency of the SPPs and its dependence on parameters of the materials and the gratings. In response to the incident light, the free electrons that form the plasma oscillate at their own natural plasma frequency; this frequency lies in the ultraviolet spectrum. The plasma’s frequency, p, can be calculated from the following equation1 p
=n
e²
0 m,
(1)
where n is the density of a free electron gas that oscillates with respect to fixed positive ions, m is the effective optical mass, e is the charge of the electron, and 0 is the permittivity of vacuum. It’s worth noting that the oscillation of the free electrons is dumping due to collisions; the characteristic collision frequency, , is equal to 100 THz at room temperature. The dielectric function of a metal, ( ), is given by1 ( )=1– where
p²
( ²+i
is the frequency of the incident radiation.
),
(2)
54
D. K. Mynbaev & V. Sukharenko
The key to the investigation of the excitation of the SPPs is the relationship between their propagation constant, , and the frequency, , of the incident light. At the interface between the metal of infinite optical thickness and a dielectric, the propagation constant can be represented as1 c²
( )
2
( ( )+ 2) = ,
(3)
where 2 is the electric permittivity of a dielectric. Fig. 3 visualizes the relationship between and . The straight lines represent lines of light. Therefore, in order to couple the SPPs, has to be greater than the propagation constant of light in the given dielectric. This necessary condition for exciting SPPs leads to the need to use grating or prism coupling. We consider gratings because they are more suitable to our application.
Fig. 3. Normalized
vs. normalized , as described by Eq. 3. (Here
silica
= 2.25 and
air
= 1.)
It’s worth noting that, in general, at the large value of propagation vectors SPPs frequency approaches characteristic surface plasmon frequency1 sp
= p / (1 +
dielectric).
(4)
WDM Demultiplexing by Using Surface Plasmon Polaritons
55
In our arrangement, the plasmonic frequency, sp, strongly depends on the incident frequency and the top grating — the confinement of the thin corrugated metal strip and the top dielectric. Similarly, the radiated frequency strongly relates to the plasmonic frequency and the bottom grating. These relationships can be seen from the following equations, where Eq. 5 governs the excitation1 and Eq. 6 — radiation4: n
/c² sin ± vg =
top
(5)
and ±n
/c² sin = ± vg ±
bottom,
(6)
where v is the order number, g = 2 / grating pitch is the grating vector, n is the refractive index of the top (incident) dielectric, is the angle of incidence, is the angle of radiation, and is the propagation constant of SPPs. Eq. 5 and Eq. 6 show the relationship among the propagation constant, the frequency of the incident light, the radiated frequency, and pitches of the top and bottom gratings. 3.
The Concept of WDM Demultiplexing with SPPs
We have investigated the concept of WDM demultiplexing by using SPPs: The transmission optical fiber is shaped into a cone to increase light intensity at the tip. The corrugated thin silver film is placed perpendicular to the tip and submerged in dielectric material on both sides with a different electric permittivity on each side. Incident light is launched onto the entire surface of the top grating; thus, SPP excitation occurs almost uniformly across the entire metal film. The bottom grating, from which SPP radiation is collected, has variable pitches; thus, the relationship between the excited SPP frequency and the radiated frequency changes across the metal film. This scheme is the subject of our discussion today. Here’s how we conducted our study: First, we investigated how the radiation frequency follows the change of incident and the bottom-grating pitch. In these calculations we assumed the top (from the top dielectric to the metal) incident angle is constant and equal to 22.20. Then we computed the angle of transmission within the metal by means of Snell’s law. We also used Snell’s law to compute the angle at which electromagnetic power radiated from the first-order bottom grating into the bottom dielectric. Another assumption was that our metal is silver whose permittivity is given by4 = -225.3189 + 1.9863*10-13 – 6.0794*10-29 2 + 8.3810*10-45 3 – 4.3004*10-61 4 + i(832575 – 1.3279*10-13 + 9.0474*10-29 2 – 3.2880*10-44 3 + 6.659*10-60 4 - 7.0893*10-76 5) (7)
56
D. K. Mynbaev & V. Sukharenko
Yet the other assumption was that is considered a difference between 0, given in Eq. 3 (a propagation constant at the confinement between the top dielectric and the optically infinite metal), and , defined by Eq. 5 (which represents propagation constant matching at the confinement between the top dielectric and metal). The more accurate value of has to incorporate the thickness of the metal strip, thus enabling us to calculate LRSPP and SRSPP at specific condition. However, if the thickness of the metal strip is 100 nm or greater, LRSPP and SRSPP have almost identical frequencies1. Thus, we assumed our metal film to be about 100 nm in thickness. Fortunately, the main idea of the proposed WDM demultiplexing scheme does not depend heavily on this assumption. Under these assumptions and determining the top grating pitch equal to 380 nm, we were able to construct Fig. 4, showing the reflection coefficient as a function of the incident frequency. In Fig. 4, Graph 1 has been built by using Eq. 9b, whereas Graph 2 has visualized Eq. 9a. rp = n2 cos
t
– n1 cos β / n2 cos
t
+ n1cos β,
(9a)
where t - angle of transmittance, β - angle of incidence, n1 - top (incident) medium refractive index, n2 - bottom (transmitted) medium refractive index.
Fig. 4. Reflection coefficient vs. incident frequency.
WDM Demultiplexing by Using Surface Plasmon Polaritons
R = 1- (4 Im[ 0 ] Im[
]) / ( – ( 0 +
))² + (Im[ 0 ] + Im[
])²
57
(9b)
One can see that at 2.3 rad/s light penetration into the top metal grating became much more pronounced, which resulted in more effective SPP excitation at this frequency. It’s quite interesting to observe that the both graphs show a drop at almost the same frequencies. The propagation constant, , given in Eq. 9b, must have excitation at frequencies where the graph drops below the 0.99 coefficient of reflection; therefore, only these frequencies will excite SPPs. At these frequencies, radiation at the bottom grating must take effect; thus, altering the grating pitch of a bottom grating will result in altering the radiated frequency. Different incident frequencies will correspond to different ’s. Imagine the entire surface of the bottom silver grating divided into sections with different pitches. Fig. 5 visualizes this effect showing a grating with two different pitches.
Fig. 5. A grating surface with different pitches.
If the first section of the grating has, for example, a pitch of 360 nm, then at an incident frequency of 2.316*1015 rad/sec with specific , the radiated frequency will be equal to 9.8*1015 rad/sec. At the same time, if the second section has a pitch of 380 nm, then at a different incident frequency of 2.275*1015rad/sec, the radiation frequency will be equal to 9.8*1015 rad/sec, exactly the same frequency as in the first section with pitch 360 nm. Fig. 6 whose graphs were built based on Eqs. 5, 6 and 9b explicitly demonstrates this phenomenon.
58
D. K. Mynbaev & V. Sukharenko
Fig. 6. Radiated frequency vs. incident frequency at various grating pitches.
Therefore, altering the pitches of the bottom grating, we can construct the metal strip, which will radiate exactly the same frequency at different locations on the bottom surface in response to different incident frequencies. Simply knowing from which section plasmonic radiation is received, we can uniquely identify the incident frequency. Practically speaking, we needed to place PDs at different points of the bottom grating and identify their positions with respect to the specific pitches. We also needed to place an optical filter between the bottom grating and the PD array that allows for the transmission only one radiated frequency, 9.8*1015 rad/s in our example. Then, identifying from which PD the signal is collected, one can find out from which incident wavelength this signal originates. Thus, this scheme works as a WDM demultiplexer at the nanometer scale. Fig. 7 illustrates the construction of this scheme.
WDM Demultiplexing by Using Surface Plasmon Polaritons
59
Fig. 7. WDM demultiplexer arranged at the tip of the coned optical transmission fiber.
The proposed scheme puts all PDs in a very comfortable regime because they constantly work at the same frequency; it can be easily arranged for this frequency to be at the maximum sensitivity of the PDs. Fig. 8 offers a demonstration of the dependence of radiated frequency on both the incident frequency and the bottom pitch. Radiated frequency also depends on the top-grating pitch, as shown in Fig. 9. At different incident pitches, the SPPs ’s at the confinement between the silica and silver will change too. This could be a quite useful phenomenon, which would allow for greater difference between the radiated frequencies, but it’s a subject for further study.
60
D. K. Mynbaev & V. Sukharenko
Fig. 8. Radiated frequency as a function of both the incident frequency and the bottom pitch at 0.99 reflection coefficient.
Fig. 9. Radiated frequency vs. top and bottom grating pitches.
WDM Demultiplexing by Using Surface Plasmon Polaritons
4.
61
Summary
We propose an optical scheme that can serve as a WDM demultiplexer in nano scale. Calculations show that realization of this scheme is quite possible. However, much work has to be done—both theoretically and experimentally—to develop this scheme from a practical standpoint. Acknowledgment This work was supported by PSC-CUNY grant 61558-00 39 and by the Emerging Scholars program of the New York City College of Technology. References 1. Stefan A. Maier, Plasmonics: Fundamentals and Applications, (Springer Science+Business Media LLC, 2007). 2. Igor Zozoulenko, Surface plasmons and their applications in electro-optical devices, Solid State Electronics, Department of Science and Technology, Linköping University, Sweden, (2006). http://www.itn.liu.se/meso-phot 3. Z. Chen, I.R. Hooper, and J.R. Sambles, Coupled surface plasmons on thin silver gratings, Journal of Optics A: Pure and Applied Optics, 10 (2008), 015007. 4. S. Wedge, I.R. Hooper, I. Sage, and W.L. Barnes, Light emission through a corrugated metal film: The role of cross-coupled surface plasmon polaritons, Physical Review B 69, 245418 (2004). 5. I.R. Hooper and J.R. Sambles, “Surface plasmon polaritons on thin-slab metal gratings,” Physical Review B 67, 235404 (2003).
This page intentionally left blank
SILICON AND GERMANIUM ON INSULATOR AND ADVANCED CMOS AND MOSHFETs
This page intentionally left blank
CONNECTING ELECTRICAL AND STRUCTURAL DIELECTRIC CHARACTERISTICS G. BERSUKER, D. VEKSLER, C. D. YOUNG, H. PARK, W. TAYLOR, P. KIRSCH and R. JAMMY SEMATECH, 2706 Montopolis Dr., Austin, TX 78741, USA [email protected] L. MORASSI, A. PADOVANI and L. LARCHER DISMI Università di Modena e Reggio Emilia and IU.NET, 42100 Reggio Emilia, Modena, Italy An attempt is made to correlate electrical measurement results to specific defects in the dielectric stacks of high-k/metal gate devices. Defect characteristics extracted from electrical data were compared to those obtained by ab initio calculations of the dielectric structures. It is demonstrated that oxygen vacancies in a variety of charge states and configurations in the interfacial SiO2 layer of the high-k gate stacks contribute to random telegraph noise signal, time-dependent dielectric breakdown, and the flatband voltage roll-off phenomenon. Keywords: random telegraph noise; dielectric breakdown; flatband roll-off.
1. Introduction Relentless device scaling challenges the traditional “empirical” approach to device characterization: increasing variability adversely affects the predictive capability of wellestablished statistical evaluation methods, while new materials and complex multicomponent gate stack structures may result in instabilities from process-generated defects (rather than stress-generated ones, which are usually associated with the time dependency of device characteristics). These factors point to a growing need to identify the nature of defects affecting the electrical characteristics of devices, specifically reliability, that would allow developing physics-based degradation models, as well as provide helpful feedback to process optimization efforts. In this study, we correlate electrical characteristics of metal/high-k gate stacks that contain a SiO2 layer at the gate dielectric/substrate interface to specific atomic defects in this interfacial layer. As has been demonstrated by electrical, physical, and modeling studies, Hf-based high-k films can modify the stoichiometry of the underlying SiO2 layer by rendering it oxygen-deficient. This leads to an increase in its dielectric constant and a higher density of fixed charges in this layer, which degrade channel carrier mobility. Here we focus on identifying specific characteristics of the oxygen vacancy defects by matching defect structural parameters extracted from electrical measurements to those obtained by ab initio calculations.
65
66
G. Bersuker et al.
By using the analysis for low frequency noise data, which takes into consideration multi-phonon relaxation processes induced by the charge trapping/detrapping in the dielectric, we demonstrate that the essential characteristics of traps in the SiO2 layer in high-k devices can be obtained. Strong dependency of the electron capture/emission times on defect relaxation energy allows extracting the latter value, which can be used as a defect identifier along with the defect energy and capture cross-section characteristics. Complementary modeling of the gate leakage current in high-k devices during electrical stress using the same approach yields characteristics of the traps in the interfacial SiO2 layer contributing to the trap-assisted tunneling process (TAT). Comparing defect energy characteristics from random telegraph signal noise (RTN) and TAT measurements to those obtained by ab initio calculations the electrically active defects can be tentatively assigned to oxygen vacancies. Based on these findings, as well as an earlier transmission electron microscopy/electron energy-loss spectroscopy (TEM/EELS) study of the elemental composition of the breakdown path, we propose that the breakdown path formation/evolution in the interfacial layer is associated with the growth of an oxygendeficient filament facilitated by the grain boundaries of the overlying high-k film. Interfacial layer defects were also found to control the so-called flatband voltage (Vfb) roll-off (R-O) phenomenon, which significantly limits the available options for metal/high-k transistor fabrication. This phenomenon describes a significant reduction of the effective work function (EWF) values of gate stacks consisting of a metal electrode, high-k dielectric, and interfacial SiO2 layer when the thickness of the latter layer is scaled down. It was determined that the oxygen vacancies at the interface with the Si substrate, generation of which occurs more effectively during high temperature processing and in thinner interfacial SiO2 layers, may acquire positive charge resulting in lower Vfb. In the subsequent sections, we discuss approaches to identify the interlayer (IL) defects contributing to RTN, time-dependent dielectric breakdown (TDDB), and flatband voltage roll-off. 2. Noise Generating Defects in High-k Gate Stacks 2.1. Analysis of Random Telegraph Signal Noise (RTN) Electrically-active defects in the gate dielectric are capable of generating noise (small fluctuations in the device output current [Fig. 1]). The phenomenon caused by a carrier in the channel hopping in and out of a single defect and resulting in the output current fluctuation between two discrete values, is called random telegraph signal noise (RTN) [1] (charge exchange with more than one trap results in a multilevel RTN). In large area devices containing sufficiently high density of such traps uniformly distributed through the dielectric thickness, overlapping signals from the individual traps form the 1/f dependency of the noise power spectral density. RTN provides information about the trap average capture and emission times, Fig. 1. Conventional RTN analysis is based on the ratio of these times and employs a detailed balance principle [2,3,4]. This analysis allows such characteristics as the trap thermal
Connecting Electrical and Structural Dielectric Characteristics
67
ionization energy and spatial position to be extracted while avoiding explicitly considering an electron trapping/detrapping mechanism. However, knowing the defect energy alone is not sufficient for elucidating the nature of the defect. To obtain additional defect characteristics, the analysis should be based on a wider set of independent experimental parameters contained in measured RTN data, specifically, the data sets of both trap capture times and emission times and their dependences on the electrical biases and temperature. For this purpose, we must describe the physical processes of electron trapping and detrapping.
Drain current [nA]
τc Hi 230
220
Low -0.3
-0.2
-0.1
τE
0.0
0.1
0.2
Time [ms]
Fig. 1. Typical two-level RTS signal in a high-k MOSFET. The low current state corresponds to the trap being filled; the high current state corresponds to the empty trap. The time periods in the low, τE, and high, τC, current states correspond to the electron emission and capture times, respectively.
Electron transitions in condensed media using the concept of multiphonon-assisted (non-elastic) processes (see, for instance, [5,6,7] and references therein) were considered for a variety of physical phenomena. The specific case of electron trapping at the bulk defect in the gate dielectric can be considered, within this approach, as a combination of two coherent processes [8]: 1) Electron tunneling from the transistor channel to the trap (Fig. 2a), with the probability
Ptun = exp(− xT / λe )
(1)
2) Rearrangement of the lattice atoms forming the trap (Fig. 2b) to accommodate an additional electron charge, with the probability [6,7]:
Prelax = e − ( 2 n + 1 ) S
n +1 n
p
(
I p 2S
n (n + 1 )
)
(2)
Here xT is the trap distance from the substrate; λ is the characteristic electron tunneling length; Ip(x) is a Bessel function of an order p; n is an equilibrium number of phonons; p = E0 / ω, with E0 being the total energy difference between the initial (trap is empty) and the final (electron in the relaxed trap) states of the system, S = Erelax / ω is the HuangRhys factor, and Erelax the energy associated with the displacements of the atoms in the dielectric (the trap relaxation energy) caused by electron trapping; ω is the characteristic phonon frequency associated with these displacements; and T is an ambient temperature.
68
G. Bersuker et al.
The second process, called the “structural relaxation,” might significantly affect the trapping dynamics. Indeed, to form a new atomic configuration around the trapped electron, the lattice atoms must shift from their initial equilibrium positions, which is, in general, associated with the system overcoming an energy barrier [5,6,8]: EB=(Erelax-E0)2/4 Erelax kT
(a)
(3)
(b)
(c)
EB
E relax
E0
e
Substrate
Substrate
Oxide
-e
∆Q Fig. 2. Schematic of the band diagram illustrating (a) electron tunneling to the trap and (b) trap relaxation caused by the electron trapping. (c) Total energy diagram, taking into account electon and phonon subsystems. Q – the generalized coordinate of the atomic displacements in the system. Solid and dashed parabolas correspond to the full system potential energy in the empty and filled traps, respectively. Erelax is the energy corresponding to displacements of the lattice atoms ∆Q. E0 is the energy loss by an electronic subsystem after a trapping event. The diagram indicates that electron trapping is associated with the system transitioning over the barrier EB.
This barrier is responsible for retarding the electron trapping (and detrapping) process (Fig. 2c). In the case of S >> p and kT >> ω,
Prelax ≈ e − E B / kT
(4)
Within the Shockley Read Hall (SRH) approximation, the capture and emission times can be expressed as follows:
τ c = (vt × n × σ 0 Ptun Prelax ) −1 ,
τ E = τ c exp(− (( ET + e ⋅ Fox ⋅ xT ) − EF ) / k ⋅ T ) ,
(5)
Here νt is an electron thermal velocity, n is electron concentration in the substrate, σ0 is the electronic component of the trap capture cross-section, EF is the Fermi energy in the substrate, Fox is the electric field in the oxide, e is the electron charge, and ET is the trap energy. A comprehensive analysis, which considers the total system energy including phonons in the dielectric along with the electronic subsystem, allows the measured values for the capture/emission times to be connected to the trap’s relaxation energy. The latter represents an important trap characteristic serving as a marker of the defect’s structure, in addition to the trap energy, location, and capture cross-section.
Connecting Electrical and Structural Dielectric Characteristics
69
2.2. Extracting defect characteristics For high-k gate stacks, the noise data were collected on small (W/L = 0.3 µm/0.1 µm) transistors with a 1 nm SiO2/3 nm HfO2/TiN gate stack, in the temperature range of 300K–345K, and capture/emission times were extracted. The dependence of the times versus the electrical bias and temperature were reproduced theoretically using Eqs. (1-5), see Fig. 3. The traps in the high-k MOSFETs under investigation were found to reside in the interfacial SiO2 layer, sandwiched between the HfO2 and Si substrate, around 0.3 nm from the substrate interface; the trap energy was ~ 3 eV, counting from the conduction band edge of the SiO2 dielectric. At such short tunneling distances, the capture and emission times are controlled by the structural relaxation process (Eq. (2)) with the relaxation energy of ~ 1.7–1.9 eV. -2
10
-3
10
-4
10
-5
10
T=300 K
-6
10 -2 10
τ C, τ E [s]
-3
10
T=330 K
-4
10
-5
10
-6
10-2 10 -3
10
T=345 K
-4
10
-5
10
-6
10
0.0
0.1
0.2
0.3
0.4
0.5
VG-Vt (V) Fig. 3. Experimental (dots) and calculated (lines) average capture (circles) and emission (squares) times for a 1 nm SiO2/3nm HfO2/TiN nFET. T = 300 K, 330 K, and 345 K. A theoretical fit is performed with a single set of trap parameters for the entire range of gate biases and temperatures. The inset shows schematically an electron capture-induced conformation of a neutral oxygen vacancy [9].
The extracted defect parameters, including the trap relaxation energy and the measure of the trap lattice distortion upon electron capture, were compared to the results of ab initio calculations of several possible configurations of oxygen vacancies in SiO2 [9]. The
70
G. Bersuker et al.
comparison allows the traps contributing to noise to be tentatively identified as neutral (V0) oxygen vacancies (converted into V- after electron trapping) in the interfacial SiO2 layer of the MOSFET gate stack. V- vacancy defects with similar characteristics were also identified from the RTS data collected on transistors with SiON/poly-Si gate stacks [10]. Note that the practical available frequency range of the noise measurement setup limits the observable spatial and relaxation energy window: the traps with lower Erelax can be observed farther away from the interface with the Si substrate (to keep the characteristic times above the low limit of the available measurement frequencies) [10]. 3. Defects Responsible for High-k Gate Stack Degradation/Breakdown Previous studies demonstrated a strong correlation between defect generation in the interfacial SiO2 layer (IL) in high-k stacks under stress (as measured by the frequencydependent charge pumping method, f-CP) [11,12] and an increase of the gate leakage current measured at low gate voltages Vg1000 C 1
2
3
4
5
6
EOT (nm) Fig. 10. Vfb roll-of (R-O) dependence on processing temperature in the 2 nm ALD HfO2 capacitors with the RuO2 gate electrode. EOT was changed by varying thickness of the underlying SiO2 layer (terraced oxide capacitor structures [22]).
The following general dependencies of R-O characteristics were observed (to various degrees) in all devices with the metal/high-k/SiO2 gate stacks: • R-O increases when the electrode has a higher WF, and when high-k film is thicker. The effect of the electrode WF on R-O can be observed by comparing high-k gate stacks with different electrodes in Fig. 9. R-O onset shifts to the thicker interfacial SiO2 in higher WF stacks and its magnitude is magnified by the high-k film thickness, as demonstrated in [22].
Connecting Electrical and Structural Dielectric Characteristics
e
75
metal High-k O3
e
A
O1
B
O2
SiO2
+
+
C
Si
Fig. 11. Schematic of the proposed the R-O model. Shaded area represents strained transitional SiO2 region.
Fig. 12. Atomistic structure of the Si/SiO2/HfO2 stack. A, B, and C are Si atoms with +1, +2, and +3 oxidation states. O1, O2, and O3 are removed oxygen atoms.
• R-O increases with a higher processing thermal budget (either higher temperatures and/or longer anneals (Fig. 10)). • R-O is slightly greater on p-type substrates. Mo2N gated devices show that the R-O could be up to 100 mV more negative on ptype than on n-type substrates [23] although the substrate effect is less in mid-gap TiN devices. 4.1. Roll-off mechanism Much of the above data points to the bottom SiO2 interfacial layer as a region in the metal/high-k stack primarily responsible for R-O while other gate stack components may modulate, to a certain degree, its magnitude. We propose [24] that the R-O phenomenon as caused by enhanced positive charge generation within the interfacial SiO2 layer when it becomes thinner than a certain critical value. These positive charges are suggested to be associated with the oxygen vacancies formed in SiO2 due to its interaction with the overlaying high-k film/metal gate stack [15]. The generation of oxygen vacancies is expected to be significantly enhanced when O atoms are consumed from the highly strained transitional region in SiO2 adjacent to the Si substrate [25]. Since a vacancy distribution through the SiO2 thickness is oxygen diffusion-limited, enhanced vacancy generation due to oxygen consumption from the strained transitional SiO2 layer is observed in only relatively thin SiO2 films, when their thickness becomes comparable with the characteristic O diffusion length under given processing conditions. A final postprocessed density gradient of O-vacancies in SiO2 is affected by the processing temperature, SiO2 density, high-k film thickness and composition, etc.
76
G. Bersuker et al.
4.2. Generation of positively charged defects Flatband Voltage (V)
IL f ix e d c h a r g e ( E 1 2 /c m 2 )
3.5
2.5
1.5
0.5
-0.5 3.5
4.5
5.5
6.5
7.5
8.5
IL dielectric constant
Fig. 13. Correlation between the dielectric constant and fixed positive charges in ~1nm interfacial SiO2.
1.2
w/ F implant Control
1.0 0.8 0.6 1
2
3 4 EOT (nm)
5
6
Fig. 14. R-O suppression by fluorine incorporated in the interfacial SiO2 layer.
As was shown [26,27], Hf-based dielectrics consumes O from the interfacial SiO2 layer due to a thermodynamic force that drives oxygen vacancies to migrate towards the SiO2 layer. Therefore, the concentration of vacancies in SiO2 is affected by their density in the adjacent high-k film, which effectively works as a source of vacancies. On the other hand, the calculations showed that for a high WF metal with a large oxide formation enthalpy, the dielectric/metal interface may be intrinsically unstable with respect to metal oxidation with simultaneous formation of oxygen vacancies in hafnia [28]. Thus, a higher density of oxygen vacancies in the interfacial SiO2 layer in a high-k gate stack with higher WF metal electrodes should be expected. To compare the formation energies of the oxygen vacancies in the SiO2 bulk and transitional layer adjacent to the Si substrate, the Si/SiO2/HfO2 structure was modeled using density functional theory within the local density approximation (a plane wave code VASP); the total energy calculations were performed using the Vanderbilt-type ultra-soft pseudopotential method [29] (Fig. 12). It was found that formation energies for the vacancies at the O2 and O3 sites are lower than that of O1 by 0.71 and 0.80 eV, respectively. For a doped substrate, two electrons left on the Si dangling bonds on the vacancy site can transfer to Si, thus forming a positively charged defect. To verify the relationship of the positive charge at the dielectric/substrate interface and oxygen content, high-k transistors with varying stoichiometry of the interfacial SiO2 layer (by changing the Si substrate treatment) were fabricated using a standard CMOS process [15]. After fabrication, all stacks were found to have approximately the same 1.1 nm SiO2 layer physical thickness (by TEM) while k values of the post-processed SiO2 layers (extracted using the EOT values of the total transistor gate stacks and of highk/metal stacks) showed a strong dependence on the initial pre-high-k deposition conditions. Since the k value of the SiO2 was shown to depend on the density of underoxidized Si ions there, the observed increase of k(SiO2) indicates the SiO2 layers in the
Connecting Electrical and Structural Dielectric Characteristics
77
corresponding gate stacks had a greater oxygen deficiency. On the other hand, by modeling the observed Vfb vs. equivalent oxide thickness (EOT) dependence obtained on these devices, the density of the positive charges in each of the stacks was extracted (Fig. 13). The data demonstrates a clear correlation between the degree of SiO2 oxygen deficiency (as reflected by its k value) and the magnitude of positive charges in the SiO2. 4.3. Origin of R-O defects and R-O suppression Experimental data support the premises of the proposed R-O mechanism. Indeed, thicker high-k films function as a stronger source of oxygen vacancies in the interfacial SiO2 layer leading to greater R-O. Similarly, higher WF electrodes, which generate oxygen vacancies in transition metal oxides more efficiently, are expected to enhance R-O by increasing the oxygen vacancy supply (through the higher-k film) to the interfacial oxide. Both generation and diffusion of oxygen vacancies in the dielectric stacks are controlled by the thermal budget leading to a strong R-O temperature dependency. Since the charge state of oxygen vacancies in the interfacial SiO2 layer depends on the position of the substrate Fermi level, a positively charged vacancy state is more probable with p-substrates. Note that the R-O phenomenon may lead to misinterpretation of the mismatch of the measured and intrinsic WF values as caused by the Fermi-level pinning. If R-O is caused by oxygen deficiency of the interfacial SiO2 layer, then passivation of the oxygen vacancies should suppress R-O. Vacancies could be passivated, for instance, by incorporating either oxygen or fluorine into the interfacial layer. Indeed, implanting F+ in the gate stack (1 nm SiO2/2 nm HfSiO) followed by annealing at 1000°C/10 sec [30] significantly reduces R-O (Fig. 14) (post-processing SIMS data shows F accumulates at the SiO2/Si interface). Similarly, R-O has been suppressed by employing a low-temperature (2000 ohm-cm) Si substrates demonstrated record DC and high speed performance of any thin film transistors. Drain current on/off ratios better than 1012 and sub-threshold voltage swing values of less than 100mV/decade could be obtained. Devices with 2µm gate lengths produced exceptionally high current densities of >750mA/mm. Shorter gate length devices (LG=1.2µm) had current and power gain cut-off frequencies, fT and fmax, of 2.9GHz and 10GHz, respectively. Keywords: Nanocrystalline; ZnO; thin films; pulsed lased deposition; atomic layer deposition; FET.
1. Introduction Thin film transistors (TFT) made from amorphous or organic semiconductors are commonly used in the control circuit of large area display electronics such as flat panel TV screens. They can also be used in applications requiring flexible or non-planar surfaces where the use of regular single crystal electronics is problematic. The usefulness of TFTs, however, has not been extended to high performance applications due to the orders of magnitude lower electronic properties of thin films compared to their single crystal counterparts. Very low electron mobilities (0.1–1 cm2/V.s) typically associated with conventional TFTs based on amorphous Si and organic semiconductors together with poor threshold voltage control issues prevented these technologies from advancing to more demanding applications.1 Nanocrystalline ZnO (nc-ZnO) thin films offer a unique solution to improving TFT performance to levels comparable to single crystal semiconductors while maintaining their thin film properties.2 High performance TFTs were fabricated using a variety of metal-oxide semiconductors such as ZnO3,4, gallium indium zinc oxide (GIZO)5, InGaO36 and zinc tin oxide.7 As a binary compound, ZnO represents the simplest form of metal-oxide semiconductors that are being developed for thin film electronics applications. Its simpler structure makes it easier to control its composition during film deposition and therefore many different types of growth techniques have been successfully implemented for thin 171
172
B. Bayraktaroglu, K. Leedy & R. Neidhard
film transistor applications, including sputtering,8,9 PLD,2,10 ALD,11,12,13 MOCVD,14 and spin coating.15 In addition to structural simplicity, ZnO also has some other interesting properties that make it particularly attractive for thin film applications. First, ZnO is a wide bandgap semiconductor (Eg=3.37eV), which makes it transparent to infrared and visible light,16 enables it to support high electric fields, and maintain low leakage current even at high temperatures. Second, it maintains most of its intrinsic single crystalline electronic properties in thin films that contain a large number of crystalline defects.17 These two properties taken together, i.e. defect tolerant operation of a wide bandgap semiconductor, allow ZnO to exploit superior electronic and optical benefits of wide bandgap semiconductors and the application diversity of thin film electronics. ZnO TFTs are currently all NMOS type devices with n-type conduction channels. As with all MOS-type FETs, ZnO TFT performance depends strongly on the properties of the ZnO and gate insulator thin films as well as the interface states between these layers. Therefore, improvements in both the ZnO and the gate insulator films must be considered together to achieve higher performance. PLD-grown nc-ZnO TFTs on Plasma Enhanced Chemical Vapor Deposited (PECVD) SiO2 gate insulators have shown drain current on/off ratios better than 1012, electron mobility of 110 cm2/V.s, and current density of higher than 400 mA/mm of gate width.18 Despite the fact that a relatively low dielectric constant gate insulator was used, the maximum device transconductance was 80 mS/mm for 2µm gate length devices. Higher dielectric constant gate insulators such as Al2O3 and HfO2 are attractive choices19,20 for improving transconductance values. In this paper, we have systematically investigated the influence of ZnO growth conditions on the device performance by examining the nanocrystalline structures of the films by transmission electron microscopy (TEM), atomic force microscopy (AFM), X-ray diffraction and various electrical measurements on the fabricated devices. Film growth temperatures in the range of 25°C to 400°C were investigated. The influence of gate insulator was also investigated using ALD-grown high dielectric constant films such as Al2O3 and HfO2. Finally, it was shown that 1.2µm gate length nc-ZnO TFTs are capable of microwave amplification at frequencies as high as 10GHz. 2. Experiment 2.1. Thin Film Preparation and Characterization 2.1.1. PLD-grown ZnO Thin Films ZnO films were deposited in a Neocera Pioneer 180 pulsed laser deposition system with a KrF excimer laser (Lambda Physik COMPex Pro 110, λ=248 nm, 10 ns pulse duration). The base pressure of the chamber was 4x10-8 Torr. ZnO films were deposited with a laser energy density of 2.6 J/cm2, laser repetition rate of 30 Hz, deposition temperature of 25°C to 400°C, and oxygen partial pressure of 10 mTorr during the deposition. The target was a 50 mm diameter by 6 mm thick sintered ZnO ceramic disk (99.999%). Additional process parameters are discussed elsewhere.21
ZnO Nanocrystalline High Performance Thin Film Transistors
173
The ZnO crystal structure was determined by using a PANalytical X’Pert Pro MRD x-ray diffractometer. The film morphology was analyzed with an FEI DB235 scanning electron microscope (SEM) and a JEOL 4000EX transmission electron microscope (TEM) operating at 400 kV. Surface roughness was measured with a Veeco Dimension 3000 atomic force microscope (AFM) and analyzed with SPIP image processing software. 2.1.2. PECVD and ALD-grown Gate Insulator Films In this study, three different gate insulators were used. SiO2 films grown in PlasmaTherm 790 PECVD system at 250°C were used for most of the devices discussed here. We have also used higher dielectric constant insulators grown by ALD. Al2O3 and HfO2 films were deposited at 250°C in a Cambridge Nanotech Fiji F200 ALD chamber. The Al2O3 depositions used trimethylaluminum and water as the aluminum and oxygen source, respectively. The HfO2 depositions used tetrakis(dimethylamido)hafnium(IV) as the hafnium source and a remote radio frequency 300W O2 plasma as the oxygen source. Argon was used as a precursor carrier gas and plasma purge gas. Nine point wafer maps of film thickness were measured with a Horiba Jobin Yvon UVISEL spectroscopic ellipsometer. Typical 20 nm films exhibited a 1σ thickness variation of < 1%. 2.1.3. TFT Fabrication and Characterization Devices for low frequency device characterization were fabricated on 200nm thick SiO2-covered Si wafers. Higher resistivity substrates (>2000 ohm.cm) were used for high frequency device fabrication to minimize capacitive parasitics. A bottom-gate configuration was used for device fabrication with a Ni/Au (5 nm/120 nm) gate below the gate insulator. Devices for low frequency characterization had multiple gate fingers, whereas high frequency devices had only two gate fingers as shown in Figure 1.
G
G S (a)
S D
S (b)
S D
Figure 1: Images of nc-ZnO TFTs (a) Multi-finger device for low frequency characterization, WG=400µm (b) High frequency device, LG=1.2µm, WG=100µm.
174
B. Bayraktaroglu, K. Leedy & R. Neidhard
Identical device fabrication sequences were employed for the fabrication of low frequency and high frequency devices. The fabrication approach was typical of bottomgate transistors and involved the fabrication of gate metal by evaporation and lift off techniques followed by the fabrication of the gate insulator and ZnO thin films over the gate metal. Based on the observation that the ZnO layer thickness in the range of 30-200 nm did not have a significant impact on the device performance; the film thickness was kept constant at 50 nm. As mentioned above, the ZnO deposition temperature ranged from 25°C to 400°C. Post-growth annealing was performed on some wafers in the temperature range of 400°C to 600°C in ceramic ovens containing clean room air. The device area was defined by mesa etching of the ZnO film in dilute HCl solution. Via holes were opened in the gate insulator layer over the gate contact pads using reactive ion etching before the fabrication of Ti/Pt/Au (20/30/350 nm) source/drain contacts by evaporation and lift of techniques. The source and drain contacts had various amounts of overlap to the gate electrode in three designs. The overlap amounts were approximately 1.5, 1, and 0.5µm, as measured by SEM, in designs designated as Design A, Design B, and Design C, respectively. No passivation layers were used over the top surface of the devices in this study. The devices were dc characterized using an Agilent 4156C Precision Semiconductor Parameter Analyzer. Microwave performance was measured at room temperature using on-wafer coplanar microwave probes in an Agilent 8364B Precision Network Analyzer. From the measured s-parameters, current gain, |h21|2, and maximum available gain (MAG) values were determined as a function of frequency. 3. Results and Discussion 3.1. Nanocrystalline ZnO Films X-ray diffraction scans of the ZnO films in Figure 2 exhibited a highly textured c-axis orientation with only the ZnO (002) peak present, consistent with other studies of PLD ZnO films.22,23,24,25 The relative intensities of the (002) peak in films deposited between 25°C and 400°C generally increased with increasing deposition temperature. All 2 positions were less than the 34.421° 2 (002) peak from the JCPDS #36-1451 powder diffraction file indicating strained lattice structures. The full-width at half-maximum value for the (002) peak decreased with increasing deposition temperature indicating improved film crystallinity. Further details on XRD results are given elsewhere.21 Cross sectional TEM images reveal densely compacted, highly faulted columnarshaped grains that predominantly extend through the thickness of the ZnO film as shown in Figure 3. At deposition temperatures of 75°C and below, diffraction contrast differences revealed elongated and, in and some areas, nearly equiaxed grain structures. Films deposited at 100°C and higher displayed a more uniform packed and elongated grain structure with more homogeneous diffraction contrast. Smooth interfaces between ZnO and SiO2 are observed at all deposition temperatures while the surfaces of the ZnO grains are predominantly dome-shaped.
ZnO Nanocrystalline High Performance Thin Film Transistors
175
Intensity (arb. units)
ZnO (002)
400°C 200°C 25°C 20
25
30
35
40
45
50
2Θ
Figure 2: X-ray diffraction patterns of ZnO thin films deposited at 25°C to 400°C.
Corresponding surface SEM images in Figure 3 show uniform surface morphologies for all ZnO depositions with no trend in grain size as deposition temperature increases. AFM images with 500 nm x 500 nm collection areas in Figure 3 also show that the films were predominantly smooth. RMS roughness values ranged from 0.65 nm to 1.65 nm. Although the lowest roughness occurred in a film deposited at 200°C, no trend in roughness as a function of deposition temperature was observed. The RMS roughness is similar to other reported values of ZnO deposited by PLD.24,25 Grain size calculations based on grain boundary intercepts in a scanning probe image processing software indicated 25 nm to 35 nm ZnO grains, again with no trend observed as a function of deposition temperature.
Figure 3: Cross sectional TEM images, surface SEM images and AFM images of ZnO thin film deposited on 20 nm SiO2/Si at 25°C, 75°C, 100°C, 200°C, 300°C and 400°C.
176
B. Bayraktaroglu, K. Leedy & R. Neidhard
3.2. Low Frequency Devices The intrinsic properties of nc-ZnO TFTs were examined using various gate length devices (2 - 25µm). The common source ID*LG vs. VD characteristics of devices with SiO2 gate insulators and various gate lengths are shown in Figure 4. In the linear region, the devices all show nearly identical characteristics indicating normal device size scaling. In the saturated region, shorter gate length devices exhibit higher current levels due to channel length modulation (CLM)26 with drain bias. This effect may be reduced by proportional decrease in tox or increase in channel doping density. The devices studied here all had tox=30nm and an estimated channel doping levels of 2x1016 cm-3. Devices were fully depleted and the I-V characteristics did not exhibit any droop due to self heating. No significant hysteresis was observed with any gate length device. 250
LG=2µm 5µm
ID*LG (mA.µ µ m)
200
25µm 150 100 50 0 0
10
20
Drain Voltage (V) Figure 4: Normalized I-V characteristics of nc-ZnO TFTs with different gate lengths.
The influence of gate length on device performance can also be observed in the transfer characteristics shown in Figure 5. As before, no significant hysteresis effects are seen as the gate bias is swept in both directions. In the sub-threshold voltage region, where the drain current increases exponentially with gate bias, all devices exhibit almost identical characteristics. The sub-threshold voltage swing value in this region is about 75mV/decade, which corresponds to an interface state density of 2.5x1011 cm-2 using relative dielectric constant of 3.9 for SiO2 and the expression developed in Ref. 27. Note that in this type of estimation, the interface states include both the surface and the bulk states and all such states are assumed to be independent of energy. A gate bias dependent electron mobility characteristics were observed with a maximum value of 110 cm2/V.s.18 The drain current on/off ratios are about 1012 for all devices. The parameter that is most impacted by the gate length is the drain current density, as expected from normal device size scaling.
ZnO Nanocrystalline High Performance Thin Film Transistors
177
1.E-01 1.E-02 1.E-03 1.E-04 1.E-05
Drain Current (A)
1.E-06 1.E-07 1.E-08
2um 5um
1.E-09
10um
1.E-10
15um 20um
1.E-11
25um
1.E-12 1.E-13 1.E-14 1.E-15 -6
-4
-2
0
2
4
6
8
10 12
Gate Voltage (V) Figure 5: ZnO TFT transfer characteristics as a function of gate length. Drain current on/off ratios better than 1012 and sub-threshold voltage swing values of 75mV/decade were obtained with all size devices.
3.3. Influence of ZnO Growth Temperature ZnO films grown at different temperatures all maintained their nanocolumnar structures, as discussed above. The electrical characteristics of the transistors made from these films also maintained their basic characteristics. All devices had excellent on/off ratios of about 1012 and no hysteresis characteristics. Threshold voltage values estimated from the linear portion of the ID vs. VG curves changed slightly from 1.05V to 0.9V as the growth temperature increased from 25°C to 400°C. The only device parameter that seems to be influenced by the growth temperature is the drain current density, which is related to the film conductivity and the electron mobility. Figure 6 shows the current density of devices with different gate lengths whose ZnO films were grown at different temperatures. The current density increases with gate length linearly for devices grown at the same temperature as seen before (see Figure 4). All device sizes showed sharp increase in current density as the growth temperature increased from 25°C to 200°C. Beyond 200°C, a slight decrease in current density was observed. Beyond 400°C, the current density rapidly drops to near zero as nanocolumns begin to be separated from each other. Based on these observations, we speculate the packing of nanocolumns is maximized at 200°C.
178
B. Bayraktaroglu, K. Leedy & R. Neidhard
Figure 6: Current density dependence on film growth temperature and device gate length. Measurements were made on devices with WG=400µm at VD=14V and VG=12V.
Drain current density is further influenced by post-growth annealing of films prior to device fabrication. Films grown even at 400°C showed significant increase in current density after annealing at 400ºC for 1hr. The current density increase was nearly 2X, 3X and 3.5X when annealed at 400ºC, 500ºC and 600ºC, respectively, as shown in Figure 7. Based on the observations that the nanocrystal size increases with post-growth annealing, (to 40nm after 600°C anneal) the current density increase can be due to higher conduction across smaller number of grain boundaries for a given size device. From Figure 7, it can be estimated that devices with 1µm and 0.5µm gate lengths made with films post-growth annealed at 600ºC will be capable of over 1.2A/mm and 2A/mm current density, respectively.
µ Figure 7: Current density dependence on post-growth annealing and device gate length. Measurements were made on devices with WG=400µm at VD=14V and VG=12V.
ZnO Nanocrystalline High Performance Thin Film Transistors
179
3.4. Influence of Gate Dielectric Higher dielectric constant gate insulators can be used to enhance the field effect and increase the device transconductance. Short channel effects observed with low dielectric constant insulators such as SiO2 (e.g. see Figure 4) can be mitigated using insulators with higher dielectric constants. We have used ALD-grown Al2O3 and HfO2 in this work and analyzed the device performance enhancements with respect to the baseline SiO2 gate insulators. Figure 8 shows the I-V characteristics of three same size TFTs with 20nm thick gate insulators and 50nm thick ZnO layers grown at 200°C. The same gate bias was applied to all devices for a direct comparison. Because of differences in breakdown strengths in the insulators studied, the maximum gate bias was limited to 10V. This limitation was set by the breakdown voltage of HfO2 at 5 x 106 V/cm. The breakdown strengths of the other insulators were about 2X higher.
! Figure 8: I-V characteristics of nc-ZnO TFTs with different gate insulators. The device size was LG=5µm, WG=400µm, tox=20nm. Gate bias was increased from 0 to 10V in 2 V/step increments.
It can be seen in this figure that all three insulators are suitable for ZnO TFT applications. The lack of hysteresis in device characteristics indicates the absence of bulk or interface traps. Higher and more saturated drain current characteristics are achieved with HfO2 and Al2O3 than SiO2 at the same gate bias values indicating that both the transconductance and short channel effects are improved. Figure 9 shows in more detail the improvements in device transconductance with the use of higher dielectric constant gate insulators. For all device sizes, from LG=2µm to 25µm, the transconductance improved by a factor of 2 and 3.5 with respect to SiO2 by the use of Al2O3 and HfO2. This improvement factor is somewhat less than the ratios of dielectric constants. For example, the HfO2/SiO2 dielectric constant ratio is about 5, whereas the transconductance improvement is only 3.5. The difference between these ratios can be explained by the presence of higher interface state density in HfO2/ZnO than in SiO2/ZnO. It is therefore
180
B. Bayraktaroglu, K. Leedy & R. Neidhard
" # "
"
$
possible that even higher transconductance values can be achieved with HfO2 gate insulators with further improvements in interface quality. The highest transconductance value achieved was 135 mS/mm for 2µm gate length devices using HfO2 gate insulator. This is substantially higher that the best value (80 mS/mm) previously obtained with SiO2 gate insulators at higher gate bias conditions.18
µ Figure 9: ZnO TFT tranconductance measured at VD=12V and VG=10V for 3 different gate insulators. LG=5µm, WG=400µm and tox=20nm.
3.5. High Frequency Devices High frequency devices fabricated on high resistivity Si substrates were examined for their microwave signal amplification potential. The gate insulator for these devices was SiO2 and the ZnO films were grown at 400°C. Three different device designs all with LG=1.2µm and WG=2x50µm, but with varying amounts of gate electrode overlap with source/drain contacts, were biased at VG=6V and VD=11V for small-signal microwave tests. From the measured s-parameters, the maximum available gain (MAG) and current gain (|h21|2) values were determined shown in Figure 10. All three device designs showed an identical current gain cut off frequency of 2.9GHz. The power gain cutoff frequencies of the same devices showed a strong dependence on the amount of gate contact overlap with source/drain contacts. Designs A, B, and C with overlap amounts of approximately 1.5, 1.0 and 0.5µm exhibited fmax values of 7.5GHz, 8.5GHz and 10GHz, respectively. The performance of Design A is similar to the previously reported value of 7.45GHz for the same size device.28 The improved performances of Designs B and C indicate the importance of minimizing parasitic overlap capacitances. The results obtained here with nc-ZnO TFTs compare favorably with the RF performance obtained (fT=180 MHz, fmax=155 MHz) with other metal oxide (InZnO) thin film transistors on glass substrates.29 To the best knowledge of the authors, these are the highest cutoff frequencies obtained with thin film transistors, and indicate the excellent potential of nc-ZnO TFTs for RF circuit applications.
ZnO Nanocrystalline High Performance Thin Film Transistors 20
"# "
15
15
! 10
10
5
5
$ 0
0
-5
Current Gain, |h21|^2 (dB)
Maximum Available Gain (dB)
20
181
-5 0.2
1 Frequency (GHz)
10
Figure 10: Small signal microwave characteristics of nc- ZnO TFTs.
4. Summary and Conclusions It was shown that ZnO thin films prepared by the pulsed laser deposition technique across a wide temperature range (25°C to 400°C) have ordered nanocolumnar structures with column diameters of about 25-30nm. The diameters of columns increase to about 40nm with post-growth annealing. Most critical device parameters including drain current on/off ratios, sub-threshold voltage swings, threshold voltages and hysteresis-free operation were shown to be independent of film growth temperature. The only parameter that had dependence on growth temperature was the current density, which exhibited a peak for films grown at 200°C. The current density also showed a significant increase with post-growth annealing, possibly due to reduction of the number of grain boundaries with annealing. A substantial increase in device transconductance was obtained with the use of ALD-grown high dielectric gate insulators (Al2O3 and HfO2) compared to SiO2. A record transconductance of 135mS/mm was obtained with 2µm gate length devices using HfO2 gate insulators. High frequency response of various device designs fabricated on Si substrates with 1.2µm gate lengths were shown to be sensitive to gate contact overlap with source and drain contacts. While the current gain cut off frequency was the same for all designs (fT=2.9GHz), the power gain cut off frequency, fmax, varied from 7.5GHz to 10GHz depending on the amount of overlap. Acknowledgements This work was supported in part by Air Force Office of Scientific Research under LRIR 07SN03COR (Dr. K. Reinhart) and Defense Advanced Research Projects Agency (Dr. J. Albrecht). The authors thank D. Tomich and J. Brown for ZnO film characterization.
182
B. Bayraktaroglu, K. Leedy & R. Neidhard
References 1. C. R. Kagan and P. Andry, Thin Film Transistors (Marcel Dekker Publishing, New York, 2003). 2. B. Bayraktaroglu and K. Leedy, ECS Transactions 16, 61-73 (2008). 3. P. F. Carcia, R. S. McLean, and M. H. Reilly, Appl. Phys. Lett. 88, 123509 (2006). 4. P. K. Shin, Y. Aya, T. Ikegami, and K. Ebihara, Thin Solid Films 516, 3767 (2008). 5. W. Lim, S. H. Kim, Y. L. Wang, J. W. Lee, D. P. Norton, S. J. Pearton, F. Ren, and I. I. Kravchenko, J. Electrochem. Soc. 155, H383 (2008). 6. H. Q. Chiang, D. Hong, C. M. Hung, R. E. Presley, and J. F. Wager, J. Vac. Sci. Technol. B 24, 2702 (2006). 7. H. Q. Chiang, J. F. Wager, R. L. Hoffman, J. Jeong, and D. A. Keszler, Appl. Phys. Lett. 86, 013503 (2005). 8. P. F. Carcia, R. S. McLean, M. H. Reilly, and G. Nunes, Jr., Appl. Phys. Lett. 82, 1117-1119 (2003). 9. D. J. Kang et.al., Thin Solid Films 475, 160-165 (2005). 10. S. Masuda, K. Kitamura. Y. Okumura, S. Miyatake, H. Tabata, and T. Kawai, J. App. Phys. 93, 1624-1630 (2003). 11. D. Mourey, D. A. Zhao, J. Sun, and T. N. Jackson, IEEE Tran. Electron Dev. 57, 530 (2010). 12. W. S. Choi, J. Soc. Inf. Display 17, 751-755 (2009). 13. K. Kopalko et. al., Phys. Stat. Sol. (c) 2, 1125-1130 (2005). 14. S. Yoshizawa, K. Nishimura, and T. Sakurai, J. Phys: Conf. Series 100, 082052 (2008). 15. B. J. Norris, J. Anderson, J. F. Wager, and D. A. Keszler, J. Phys. D: Appl. Phys. 36, L105L107 (2003). 16. Ü. Özgür, Y. I. Alivov, C. Liu, A. Teke, M. A. Reshchikov, S. Do an, V. Avrutin, S. J. Cho, and H. Morkoç, Appl. Phys. Rev. 98, 041301 (2005). 17. K. Nomura, H. Ohta, A. Takagi, T. Kamiya, M. Hirano, and H. Hosono, Nature 432, 488-492 (2004). 18. B. Bayraktaroglu, K. Leedy, and R. Neidhard, IEEE Electron Dev. Lett. 29, 1024-1026 (2008). 19. N. C. Su, S. W. Wang, and A. Chin, Electrochem. Solid-State Lett. 13, H8 (2010). 20. P. F. Carcia, R. S. McLean, and M. H. Reilly, Appl. Phys. Lett. 88, 123509 (2006). 21. B. Bayraktaroglu, K. Leedy, and R. Neidhard, Mater. Res. Soc. Symp. Proc. 1201, H09-07 (2010). 22. L. Bentes, R. Ayouchi, C. Santos, R. Schwarz, P. Sanguino, O. Conde, M. Peres, T. Monteiro, and O. Teodoro, Superlattices and Microstructures 42, 152 (2007). 23. S. Amirhaghi, V. Craciun, D. Craciun, J. Elders, and I. W. Boyd, Microelectronics Engineering 25, 321 (1994). 24. L. Han, F. Mei, C. Liu, C. Pedro, and E. Alves, Physica E 40, 699 (2008). 25. C.-F. Yu, C.-W. Sung, S.-H. Chen, and S.-J. Sun, Appl. Surface Sci. 256, 792 (2009). 26. J. J. Liou, A. Ortiz-Conde, and F. Garcia-Sanchez, Analysis and Design of MOSFETs: Modeling, Simulation, and Parameter Extraction (Kluwer Academic Publishers, Boston, 1998). 27. A. Roland, J. Richard, J. P. Kleider, and D. Mencaraglia, J. Electrochem. Soc. 140, 3679 (1993). 28. B. Bayraktaroglu, K. Leedy, and R. Neidhard, IEEE Electron Dev. Lett. 30, 946 (2009). 29. Y. L. Wang, L. N. Covert, T. J. Anderson, W. Lim, J. Lin, S. J. Pearton, D. P. Norton, J. M. Zavada, and F. Ren, Electrochem. Sol. State Lett. 11, H60 (2008).
ZINC OXIDE NANOPARTICLES FOR ULTRAVIOLET PHOTODETECTION SHAYLA SAWYER*, LIQIAO QIN and CHRISTOPHER SHING Electrical, Computer, and Systems Engineering Department, Rensselaer Polytechnic Institute, 110 8th Street, Troy, NY 12180, United States *[email protected] Zinc Oxide (ZnO) nanoparticles were created by a top-down wet-chemistry synthesis process (ZnOA) and then coated with polyvinyl-alcohol (PVA) (ZnO-U). In ZnO-U, strong UV emission was apparent while the parasitic green emission, which normally appears in ZnO suspensions, was suppressed. A standard lift-off process via e-beam lithography was used to fabricate a detector by evaporating Aluminum (Al) as ohmic electrodes on the ZnO nanoparticle film. Photoconductivity experiments showed that linear current-voltage response were achieved and the ZnO-U nanoparticles based detector had a ratio of UV photo-generated current more than 5 times better than that of the ZnO-A based detector. In addition, non-linear current-voltage responses were observed when interdigitated finger Gold (Au) contacts were deposited on ZnO-U. The UV generated current to dark current ratios were between 4 and 7 orders of magnitude, showing better performance than the photodetector with Al contacts. ZnO-U were also deposited on Gallium Nitride (GaN) and Aluminum Gallium Nitride (AlGaN) substrates to create spectrally selective photodetectors. The responsivity of detector based on AlGaN is twice that of commercial UV enhanced Silicon photodiodes. These results confirmed that ZnO nanoparticles coating with PVA is a good material for small-signal, visible blind, and wavelength selective UV detection. Keywords: ZnO; photodetectors; nanoparticle; ultraviolet; green photoluminescence; surface passivation.
1. Introduction Zinc oxide (ZnO) is the emerging alternative semiconductor material to Gallium Nitride (GaN) for optoelectronic applications1. Its properties are close to that of GaN material with the additional advantages of lower cost, large-area native substrates, low temperature growth, and an exciton binding energy (60 meV) twice that of GaN (28 meV)2-6. The band gap of ZnO (~3.2-3.4 eV) is ideal for light-emitting diodes (LEDs), laser diodes, and photodetectors in the ultraviolet (UV) wavelength range. Its primary hindrance from greater prevalence is the lack of a stable and reproducible p-type material although current research with promising results have rejuvenated overall research interest7. Semiconductor photodetectors convert incident light within a range of wavelengths into detectable current based on their material properties. This wavelength range is tuned by the bandgap of the material. Currently, light detection in the blue/UV region is facilitated by Silicon (Si) photodetectors. At room temperature, the bandgap energy of Si (1.2 eV) is far below ideal for detection in the blue/UV region, which greatly reduces responsivity. Wider bandgap materials, including ZnO and AlGaN/GaN, are better matched in energy from the UV to blue wavelength region, resulting in an overall 183
184
S. Sawyer, L. Qin & C. Shing
increase in photosensitivity to UV relative to Si. The efficiency and sensitivity of the conversion of incident radiation to current is dominated by limitations of area, light coupling, and wavelength specificity. Currently, many nanostructures of ZnO have been studied extensively due to their quantum confinement effects which correspond to continuous tuning of the spectral wavelength and improved device performance8-16. Low cost, large area, wavelength tunable, ultraviolet photodetectors are possible with the inherent properties of ZnO nanoparticles. 2. ZnO Background The first studies of ZnO as a semiconductor material began in 1935 by C.W. Bunn with an investigation of lattice parameters. ZnO has three kinds of crystal structures, rocksalt, zinc blende and wurtzite, which are shown in Figure 1. In ambient conditions, the wurtzite structure is thermodynamically stable, so it is the most common structure of ZnO. Its hexagonal lattice is defined by two interconnected sublattices of Zn2+ and O2-. The Zn2+ ion is surrounded by the O2- ion and vice-versa, therefore, ZnO has polar surfaces with inherent piezoelectric and spontaneous polarization effects. The hexagonal lattice parameters are a = 3.25 , c = 5.2069 , and c/a = 1.60. ZnO material is intrinsically n-type. While doping with donors, Aluminum, Gallium and Indium (Al, Ga, and In) up to 1020 cm-3 is readily achieved, acceptor doping has proven to be difficult7. Common acceptors include Nitrogen, Phosphorus, Arsenic and Antimony (N, P, As, and Sb) with up to 1015 cm-3 active dopants. This is the primary limiting factor for ZnO devices to date.
Fig. 1. Stick and ball representation of ZnO crystal structure; black and gray spheres denote Zn and O atoms respectively. [Reprinted with permission from U. Özgür et al., J. of Appl. Phys., 98, 041301 (2005).]
ZnO Nanoparticles for Ultraviolet Photodetection
185
Bandgap engineering is available by adding Cadmium (Cd) or Magnesium (Mg) content to ZnO. The alloys Zn(1-x)MgxO (3.3 - 4 eV) and Zn(1-y)CdyO (2.9 - 3.37 eV) enable adjustments in optical properties with little variation of the semiconductor lattice parameter leading to heterojunctions with reduced strain. ZnO nanomaterials have been studied for the benefits associated with quantum confinement including reduced dark noise, increased absorption efficiency and the potential for large area and lower cost devices. The nanostructures, however; often exhibit a strong, parasitic green photoluminescence caused by excess Zn2+ ions and oxygen deficiency10,14,15,17. Solutions to these problems include developing effective surface modification, annealing treatment and producing stable ZnO QDs that are unagglomerated, highly crystalline, and monodisperse18-22. The ability to create reproducible Ohmic or Schottky contacts is also associated with oxygen vacancies23. These vacancies often dominate which tend to pin the ZnO Fermi level to the defect level, hindering the expected barrier producing relationship between the contact metal and semiconductor. The following sections will summarize and compare device results from bottom-up nanoparticle growth methods and an inexpensive alternative; the top down wet chemical etch method. The latter method addresses the issues associated with defect levels that often hinder performance. 3. Top-Down Wet-Chemistry ZnO Nanoparticle based UV Photodetectors Wet chemical synthesis is both inexpensive and simple. Much of the wet-chemistry research can be classified as bottom-up synthesis. They are based on a range of precursors and synthesis conditions, such as temperature, time, and concentration of reactants, leading to different sizes and geometries of the resulting particles. However, several significant challenges remain, including difficulty in preparation of a stable dispersion, poor uniformity in the coating process, and poor conductivity of the ZnO layers. The top-down approach was first reported by Sharma et al. in which bulk ZnO material was reduced to generate controllable and stable dispersions of ZnO nanoparticles through wet-chemistry24. This method enables an additional surface coating step with relative ease and uniformity. To investigate the effects of surface coating, the nanoparticles were coated with polyvinyl alcohol (PVA) (ZnO-U) and compared to uncoated ZnO nanoparticles (ZnO-A) produced by the same process. Figure 2 is a high resolution SEM image of PVA coated ZnO. The sizes range from 10 nm to 150 nm with an average value of 80 nm. The uncoated ZnO have a similar result.
186
S. Sawyer, L. Qin & C. Shing
Fig. 2. High resolution SEM result of PVA coated ZnO (ZnO-U) created by top-down wet-chemical synthesis with an average size of 80 nm.
3.1. Parasitic green photoluminescence reduction and enhanced UV absorption Typical photoluminescence results of ZnO include UV emission centered near 380 nm and a dominant parasitic green emission peak. Gong et al. investigated the optical properties of ZnO spherical nanoparticles prepared by bottom-up synthesis by varying precursor mixtures, 1-D octadecene, triocylamine and trioctylphosphine, which modify the density of oxygen near the surfaces25. Borgohain and Mahammuni encapsulated ZnO nanoparticles with passivating agents such as polyvinyl prrolidone (PVP), propionic acid (PA) and tetra octyl ammonium bromide (TOAB) also grown through bottom-up methods17. The PVA coated top-down wet chemical etch produced nanoparticles derived from bulk ZnO26. Both uncoated and PVA coated ZnO nanoparticles were spin-cast on quartz and measured by photoluminescence. Figure 3 demonstrates the effectiveness of PVA coating to suppress parasitic green emission associated with defects in the material by acting as a surface passivation layer. The deep level surface traps are responsible for hindering bandto-band recombination. PVA has a strong interaction with excess Zn2+ ions during the top-down wet-chemistry process thereby reducing trapped carriers and enhancing the proportion of carriers contributing to band-to-band emission. 3.2. Photoconductive and Metal-Semiconductor-Metal (MSM) device response Ohmic and Schottky contacts are ideally dependent upon the relative work functions of the contact metal and semiconductor. Metals for ohmic contacts with n-type ZnO include Aluminum (Al) and Indium (In) in addition to various metal combinations such as Titanium/Gold (Ti/Au), Aluminum/Platinum (Al/Pt)27. Candidates for Schottky contacts are Platinum (Pt), Gold (Au), Nickel (Ni), and Silver (Ag). Nevertheless, the surface defect states, residual surface contamination, and/or the interfacial gap between the metal and semiconductor often dominate creating some unpredictability in current-voltage (I-V)
ZnO Nanoparticles for Ultraviolet Photodetection
187
characteristics. For example, Jun et al. reported ohmic behaviour or linear I-V response with Au metal deposition on the n-type ZnO nanoparticles. Typically a non-linear response is expected but surface impurities and defects of ZnO are cited8,28-33.
Fig. 3. Emission spectra of ZnO coated with PVA (U_Em) and uncoated (A_Em) dispersed in ethanol and then spin-cast on quartz. Strong UV emission peak at 377 nm from the band edge of emission. PVA coating nearly eliminates parasitic effect. The inset is the normalized spectra.
The effects of surface passivation also contribute to I-V response. The general mechanism of photocurrent for ZnO nanoparticles begins with adsorbed oxygen molecules on the surface as negatively charged ions, capturing free electrons [O2(g) + O2-(ad)] producing a depletion layer. Incident light with energy higher than the ebandgap creates electron and hole pairs. The generated holes near the surface, neutralize chemisorbed oxygen [h+ + O2-(ad) O2(g)]. The depletion region narrows and photocurrent increases. Traps due to surface defects hinder the conducting mechanism. To investigate the photoconductivity of ZnO nanoparticles and verify the effect of surface passivation, photodetectors based on ZnO-A and ZnO-U were made respectively. A schematic diagram of the detectors is shown in the inset of Figure 4. The typical I-V plots of detectors with Al contacts above both ZnO-A and ZnO-U nanoparticle films in the dark and under 340 nm UV LED illumination with the same intensity (45.58mW/cm2) are shown in Figure 4. The linear I-V plots demonstrate photoconductive detectors based on ZnO nanoparticles. The contacts were made by depositing two irregular 100 nm Al contacts via e-beam lithography and standard lift-off process. These measurements were performed at room temperature in air. The UV photo-generated current to dark current
188
S. Sawyer, L. Qin & C. Shing
ratio (on/off ratio) in detector based on ZnO-A is about 9×103 when the bias is 20V, while for detector based on ZnO-U, the ratio is as high as 5×104, more than 5 times of that based on ZnO-A. These results are due to the higher defect concentration in ZnO-A compared to ZnO-U. Most of the photo-generated carriers are trapped by the defects during their transport to the terminals. However, the dark currents of both detectors are about 100 pA at 20V bias, which did not change with the PVA coating. This indicates PVA does not change the conductivity of ZnO nanoparticles, but acts as an effective surface passivation material to improve the photosensitivity of ZnO nanoparticles. In comparison to ZnO thin film photodetectors, the dark current of the above nanoparticle based detectors are 3 orders of magnitude less than the thin film ZnO photodetectors34-36. Jandow et al., Jiang et al. and Liu et al, demonstrated 1, 2, and 4 orders of magnitude between UV photogenerated current to dark current respectively34-36.
Fig. 4. Linear I-V response with Al contacts of (Top) Uncoated ZnO (ZnO-A) with UV/dark ratio of 9×103 at 20V bias (Bottom) PVA coated (ZnO-U) with UV/dark ratio of 5x104 at 20V bias.
ZnO Nanoparticles for Ultraviolet Photodetection
189
Fig. 5. PVA coated (ZnO-U) with Gold contacts. Non-linear I-V response with UV/dark ratio of 1x106 at 20V bias.
Figure 5 shows a non-linear I-V response of Au contacts deposited through optical lithography methods on PVA coated ZnO nanoparticles (ZnO-U) on quartz. Interdigitated finger contact arrays were designed with varying finger spacing, finger lengths, and finger widths. Each tested device demonstrated an MSM, non-linear response similar to Figure 5. When the detector with Au contacts was illuminated by a 340nm UV LED (45.58mW/cm2), the UV generated current to dark current ratios were between 4 and 7 orders of magnitude. In contrast to detectors with Al contacts deposited by e-beam through shadow mask, the UV generated current surpassed 100 uA at 20V bias for most MSM detectors which is an improvement of two orders of magnitude, and the dark current is about one order higher. Au contacts deposited by optical lithography rather than through a shadow mask via e-beam lithography may have reduced distance between contacts, thus decreasing the recombination possibility, which results in a higher UV generated current to dark current ratio. The photoresponse of the ZnO nanostructures are on the order of seconds or tens of seconds in literature. Jun et al. reported rise and decay times of 48s and 0.9s respectively for 20-150 nm solution processed nanoparticles8. Soci et al. reported rise and decay times of 23s and 33s respectively for 150-300 nm diameter nanowires created by chemical vapor deposition33. Kind et al. found 1s rise and decay times for 20-300 nm diameter nanowires31. The time response results for PVA coated ZnO nanoparticles with irregular Al contacts are shown in Figure 6. The samples were illuminated by a 340 nm LED with the intensity of 4.58mW/cm2. The results of 22s rise and 11s decay times are consistent with
190
S. Sawyer, L. Qin & C. Shing
that of nanostructures with UV generated current to dark current ratios near the same order of magnitude. The rise and decay of the photocurrent is related to the adsorption of oxidizing molecules in ambient air and the time required for oxygen molecule diffusion. Similar measurements will be explored for the samples with Au MSM contact structures.
Fig. 6. Photocurrent time response measurement with 340 nm LED excitation.
3.3. ZnO nanoparticles deposited on GaN-based substrates Wavelength selective photodetctors in the UV range were created with PVA coated ZnO nanoparticles deposited on GaN and AlGaN epitaxial substrates grown by Sensor Electronic Technology Inc (SET Inc.). The spectral response of these detectors is shown in Figure 7. When light penetrates the detector through the substrate or backside of the detector, a bandpass response related to the cut-off wavelength of substrate and ZnO nanoparticles is achieved. As shown in Figure 7, the absorption cut-off wavelengths of AlGaN and GaN are 300nm and 360nm, respectively through engineered epitaxial layer growth. When the wavelength of the light is shorter than the cut-off wavelengths, the light is absorbed by the substrate material and electron and hole pairs are generated. The photogenerated carriers do not have a conduction path to the contacts that are deposited on ZnO nanoparticles, thus current is not generated. Wavelengths longer than the substrate cutoff, penetrates the substrate and reaches the ZnO active layer which forms a MSM photodetector. If the wavelength is longer than the cut-off wavelength of ZnO, the light is not absorbed by ZnO and current is not generated. This is a bandpass response. The expected normal short pass response is also shown in Figure 7 when the light penetrates from the front side. In this case, the calculated responsivity of the ZnO nanoparticle-AlGaN substrate device is 0.8A/W at 375nm, and the spectral response is twice that of a commercial Si UV enhanced photodetector measured in the same system.
ZnO Nanoparticles for Ultraviolet Photodetection
191
Wavelength selective detectors can be tuned by the material properties of both the substrate and the nanoparticles. 4. Summary ZnO is an emerging material for blue and UV optoelectronics. Its properties are similar to GaN materials but differ in its low cost, reduced lattice strain from layered deposition of its alloys, larger excitonic binding energy, and availability of large-area native substrates. However, its prevalence is hindered by the lack of a reliable and reproducible p-type material. ZnO nanostructures introduce benefits associated with quantum confinement including reduced dark noise, increased absorption efficiency and the potential for large area and lower cost devices. High quality nanoparticles were created with a top-down wet-chemical process and coated with PVA. As a result, parasitic green photoluminescence, caused by deep level surface traps, was nearly eliminated. PVA provides surface passivation by reducing the number of excess Zn2+ ions. Photoconductive devices were created with both PVA coated and uncoated ZnO nanoparticles on quartz with Al contacts. A linear I-V response was achieved as expected. The PVA coated ZnO nanoparticles had a UV generated current to dark current ratio 5 times higher than the uncoated ZnO nanoparticles. The transient photoresponse from 340 nm LED excitation showed a rise and decay time of 22s and 11s respectively. These values are consistent with literature. Metal-semiconductor-metal devices were created by depositing Au interdigitated finger contacts through optical lithography. The UV generated current to dark ratio increased to seven orders of magnitude. Finally, wavelength selective devices were created by depositing PVA coated ZnO nanoparticles on epitazially grown AlGaN and GaN substrates. The absorption cutoff of AlGaN (300 nm) and GaN (360 nm) created a bandpass response when illuminated from through the substrate or backside of the detector. The responsivity
Fig. 7. Front and back illuminated UV detectors (left) ZnO nanoparticles on AlGaN substrate, shortpass (black line) and bandpass (red line: 300-380 nm) responsivity as it relates to the absorption of AlGaN (blue line) (right) ZnO nanoparticles on GaN substrate, shortpass (black line), and bandpass (red line: 360-380 nm), as it relates to the absorption of GaN (green line).
192
S. Sawyer, L. Qin & C. Shing
of the device with an AlGaN substrate was 0.8 A/W at 375 nm twice that of a commercial UV enhanced Si photodetector. These results indicate that PVA coating of ZnO nanoparticles during top-down wet-chemistry synthesis could be applied in creating low cost, sensitive, visible blind, and wavelength selective UV photodetectors. Acknowledgments The authors gratefully acknowledge support from NSF Industry/University Cooperative Research Center Connection One. We’d like to thank Professor Partha Dutta in Electrical, Computer and Systems Engineering Department Rensselaer Polytechnic Institute and Sensor Electronic Technology Inc. Finally, we would like to thank National Security Technologies for their support. References 1. V. A. Karpina, V. I. Lazorenko, C. V. Lashkarev, V. D. Dobrowolski, L. I. Kopylova, V. A. Baturin, S. A. Pustovoytov, A. J. Karpenko, S. A. Eremin, P. M. Lytvyn, V. P. Ovsyannikov, and E. A. Mazurenko, Zinc oxide - analogue of GaN with new perspective possibilities, Crystal Research and Technology, 39, 980–992, (2004). 2. Z. Bi, J. Zhang, X. Bian, D. Wang, X. Zhang, W. Zhang, and X. Hou, A high-performance ultraviolet photoconductive detector based on a ZnO film grown by RF sputtering. Journal of Electronic Materials, 37, 760–763, (2008). 3. G. Cheng, Z. Li, S. Wang, G. Gong, K. Cheng, X. Jiang, S. Zhou, Z. Du, T. Cui, and G. Zou, The unsaturated photocurrent controlled by two-dimensional barrier geometry of a single ZnO nanowire schottky photodiode, Applied Physics Letters, 93, 123103.1–123103.3, (2008). 4. Y. Jin, J. Wang, B. Sun, J. Blakesley, and N. Greenham, Solution-processed ultraviolet photodetectors based on colloidal ZnO nanoparticles, Nano Letters, 8, 1649–1653, (2008). 5. Y. Lin, C. Chen, W. Yen, W. Su, C. Ku, and J. Wu, Near-ultraviolet photodetector based on hybrid polymer/zinc oxide nanorods by low-temperature solution processes, Applied Physics Letters, 92, 233301.1–23301.3, (2008). 6. S. J. Young, L. W. Ji, S. J. Chang, S. H. Liang, K. T. Lam, T. H. Fang, K. J. Chen, X. L. Du, and Q. K. Xu, ZnO-based MIS photodetectors, Sensors and Actuators A: Physical, 141, 225–229, (2008). 7. D. C. Look, Zinc Oxide Bulk, Thin Films and Nanostructures, Elsevier, 21–42, (2006). 8. J. Jun, H. Seong, K. Cho, B. Moon, and S. Kim, Ultraviolet photodectors based on ZnO nanoparticles, Ceramics International, 35, 2797–2801, (2009). 9. S. Lee, Y. Jeong, S. Jeong, J. Lee, M. Jeon, and J. Moon, Solution-processed ZnO nanoparticle-based semiconductor oxide thin-film transistors, Superlattice Microstructures, 44, 761–769, (2008). 10. S. Monticone, R. Tufeu, and A. Kanaev, Complex nature of the UV and visible fluorescence of colloidal ZnO nanoparticles, Journal of Physical Chemistry B., 102, 2854–2862, (1998). 11. E. Neshataeva, T. Kummel, G. Bacher, and A. Ebbers, All-inorganic light emitting device based on ZnO nanoparticles, Applied Physics Letters, 94, 091115.1–091115.3, (2009). 12. Y. Qin, X. Wang, and Z. Wang, Microfibre-nanowire hybrid structure for energy scavenging, Nature, 45, 809–813, (2008). 13. M. Wang, Y. Lian, and X. Wang, PPV/PVA/ZnO nanocomposite prepared by complex precursor method and its photovoltaic application, Current Applied Physics, 9, 189–194, (2009).
ZnO Nanoparticles for Ultraviolet Photodetection
193
14. L. Wu, Y. Wu, X. Pan, and F. Kong, Synthesis of ZnO nanorod and the annealing effect on its photoluminescence property, Optical Materials, 28, 418–422, (2006). 15. Y. Wu, A. Tok, F. Boey, X. Zeng, and X. Zhang, Surface modification of ZnO nanocrystals. Applied Surface Science, 253, 5473–5478, (2007). 16. X. Zhua, I. Yurib, X. Gana, I. Suzukib, and G. Lia, Electrochemical study of the effect of nano-zinc oxide of microperoxidase and its application to more sensitive hydrogen peroxide biosensor preparation, Biosensors and Bioelectronics, 22, 1600–1604, (2007). 17. K. Borgohain and S. Mahamuni, Luminescence behavior of chemically grown ZnO quantum dots, Semiconductor Science and Technology, 13, 1154–1157, (1998). 18. L. Guo, S. Yang, C. Yang, P. Yu, J. Wang, W. Ge, and G. Wong, Highly monodisperse polymer-capped ZnO nanoparticles: Preparation and optical properties, Applied Physics Letters, 76, 2901–2903, (2000). 19. K. Kim, N. Koguchi, Y. Ok, T. Seong, and S. Park, Fabrication of ZnO quantum dots embedded in an amorphous oxide layer, Applied Physics Letters, 84, 3810–3812, (2004). 20. S. Mridha, M. Nandi, A. Bhaumik, and D. Basak, A novel and simple approach to enhance ultraviolet photosensitivity: activate carbon-assisted growth of ZnO nanoparticles, Nanotechnology, 19, 275705, (2008). 21. H. Xiong, Z. Wang, and Y. Xia, Polymerization initiated by inherent free radicals on nanoparticle surfaces: a simple method of obtaining ultrastable (ZnO) polymer core-shell nanoparticles with strong blue fluorescence, Advanced Materials, 18, 748–751, (2006). 22. S. Yang and C. Park, Facile preparation of monodisperse ZnO quantum dots with high quality photoluminescence characteristics, Nanotechnology, 19, 035609.1–035609.4, (2008). 23. M. W. Allen and S. M. Durbin, Influence of oxygen vacancies on schottky contacts to ZnO. Applied Physics Letters, 92, 122110–122113, (2008). 24. S. Sharma, A. Tran, O. Nalamasu, and P. Dutta, Spin coated ZnO thin films using ZnO nano-colloid, Journal of Electronic Materials, 35, 1237–1240, (2006). 25. Y. Gong, T. Andelman, G. F. Neumark, S. O’Brien, and I. L. Kuskovsky, Origin of defectrelated green emission from ZnO nanoparticles: effect of surface modification, Nanoscale Research Letters, 2, 297–302, (2007). 26. L. Qin, C. Shing, and S. Sawyer, Ultraviolet photodetection based on ZnO colloidal nanoparticles made by top-down wet-chemistry synthesis process, accepted Symposium on Photonics and Optoelectronics, (2010). 27. J. H. Lim and S. J. Park, Zinc Oxide Bulk, Thin Films and Nanostructures, Elsevier, 267–293, (2006). 28. S. E. Ahn, J. S. Lee, H. Kim, S. Kim, B. H. Kang, and K. H. Kim, Photoresponse of sol-gelsynthesized ZnO nanorods, Applied Physics Letters, 84, 5022–5024, (2004). 29. B. J. Coppa, C. C. Fulton, S. M. Kiesel, R. F. Davis, C. Pandarinath, J. E. Burnette, R. J. Nemanich, and D. J. Smith, Structural, microstructure, and electrical properties of gold films and schottky contacts on remote place-cleaned, n-type ZnO 0 0 0 1 surfaces, Journal of Applied Physics, 97, 103517.1–103517.13, (2005). 30. S. S. Hullavarad, N. V. Hullavarad, M. Mooers, and P. C. Karulkar, Fabrication of nanostructured ZnO UV sensor, In Materials Research Society Symposium Proceedings, 951, pages 216–220, (2007). 31. H. Kind, H. Yan, B. Messer, M. Law, and P. Yang, Nanowire ultraviolet photodetectors and optical switches, Advance Materials, 14, 158–160, (2002). 32. S. Liang, H. Sheng, Y. Liu, Z. Huo, Y. Lu, and H. Shen, ZnO schottky ultraviolet photodetectors, Journal of Crystal Growth, 225, 110–113, (2001). 33. C. Soci, A. Zhang, B. Xiang, S. A. Dayeh, D. P. R. Aplin, J. Park, X. Y. Bao, Y. H. Lo, and D. Wang, ZnO nanowire UV photodetectors with high internal gain, Nano Letters, 7, 1003–1009, (2007).
194
S. Sawyer, L. Qin & C. Shing
34. N. N. Jandow, K. A. Ibrahim, H. A. Hassan, S. M. Thahab, and O. S. Hamad, The electrical properties of ZnO MSM photodetector with Pt contact electrodes on PPC plastic, Journal of Electron Devices, 7, 225–229, (2010). 35. D. Jiang, J. Zhang, Y. Lu, K. Liu, D. Zhao, Z. Zhang, D. Shen, and X. Fan, Ultraviolet schottky detector based on epitaxial ZnO thin film, Solid-State Electronics, 52, 679–682, (2008). 36. Y. Liu, G. R. Gorla, S. Liang, N. Emanetoglu, Y. Lu, H. Shen, and M. Wraback, Ultraviolet detectors based on epitaxial ZnO films grown by MOCVD, Journal of Electronic Materials, 29, 69–74, (2000).
CARBON-BASED NANOELECTROMECHANICAL DEVICES STEFAN BENGTSSON, PETER ENOKSSON, FARZAN A. GHAVANINI, KLAS ENGSTRÖM, PER LUNDGREN Department of Microtechnology and Nanoscience, Chalmers University of Technology, 412 96 Göteborg, Sweden [email protected] ELEANOR E. B. CAMPBELL, JOHAN EK-WEIS School of Chemistry, University of Edinburgh, West Mains Road, Edinburgh, Scotland EH9 3JJ [email protected] NIKLAS OLOFSSON, ANDERS ERIKSSON Department of Physics, University of Gothenburg, 412 96 Göteborg, Sweden [email protected] Carbon-based nanoelectromechanical devices are approaching applications in electronics. Switches based on individual carbon nanotubes deliver record low off-state leakage currents. Arrays of vertically aligned carbon nanotubes or nanofibers can be fabricated to constitute varactors. Very porous, low density arrays of quasi-vertically aligned arrays of carbon nanotubes behave mechanically as a single unit with very unusual material properties. Keywords: carbon nanotubes; carbon nanofibers; nanoelectromechanical systems; switch; varactor.
1. Introduction The theoretical model predictions and experimental observations of actual response to electrical and mechanical stimuli of the carbon nanotube material family show that it is worthwhile to design and fabricate electronic devices using such materials. Advances for employing carbon nanotubes (CNTs) as conductors of current and/or heat in future integrated electronic systems have been extensively reported on and progress in that area has been rapid1-3. CNTs are also investigated in the role of acting as the active element in transistors4,5. With a broad perspective on electronics, carbon nanostructures have been demonstrated as an attractive choice as electrode material in sensor applications6. A different class of devices exploits the interplay between electrical and mechanical effects as the very key to the device functionality in a nanoelectromechanical system (NEMS). Although it is possible to conceive of many different devices and applications of NEMS made up of CNTs or similar materials, we focus our description on switches and varactors as model examples of such devices in this overview of recent advances towards the realization of carbon-based NEMS.
195
196
S. Bengtsson et al.
Fig. 1. Schematic pictures of the (a) singly clamped and (b) doubly clamped configurations for a beam switch. From Yousif10 reprinted with permission from IOP Publishing.
2. Modeling By importing theoretically predicted7 or measured values for the Young’s modulus of CNTs8 or carbon nanofibers (CNFs)9 and by employing continuum mechanics in conjunction with analytical expressions or boundary element methodology we can couple the mechanical actuation to voltage induced electrical forces. In this way it is possible to make predictions of the behavior of electronic devices like switches and varactors. In the case of a switch, the basic functionality is to change conductivity from infinite to zero as fast as possible and with minimum cost of energy. One attractive feature of using CNT electromechanics to realize a switching device is that the small dimensions and high stiffness will give a high resonance frequency. This enables fast switching. Furthermore,
Fig. 2. The dependence of the threshold voltage on the length of the switching beam. From Yousif10 reprinted with permission from IOP Publishing.
Carbon-Based Nanoelectromechanical Devices
197
Fig. 3. One-dimensional lumped model of the CNF varactor.
the disjunct off-state of such a device will yield a leakage current that will give negligible contribution to the total power consumption of the device10. In the case of varactors the functionality is more complex than just achieving a transition from an on-state to an off-state. The geometry considered in this paper employs a pair of nanoelectromechanical electrodes to obtain the voltage dependent capacitance characteristic of varactors. These non-linear circuit elements find use e. g. in voltage controlled oscillators, and some of the critical features of the varactors are their capacitance per unit chip area and their swing in capacitance for the applicable voltage range. The modeling methods employed to describe these devices comprise boundary element calculations to resolve the geometrical effects on the nanostructures as well as simplistic analytical descriptions which can capture the qualitative behavior and give first order estimates of critical parameters. 2.1. Switches In order to evaluate the optimal performance of digital switching devices fabricated using carbon nanotubes, a continuum mechanics approximation has been employed to describe mechanically switching beams of either multi-wall carbon nanotubes (MWCNT) or dense bundles of single-wall carbon nanotubes (SWCNT)10. The thickness of the beam, t, is a crucial design parameter among the geometrical dimensions where the beam length, L, and the nominal distance between beam and the actuating electrode, g0, also play an important role. Figure 1 shows a schematic illustration of the geometrical configurations considered in the paper by Yousif et al.10 and in Figure 2 the calculated threshold voltage to turn the switch on is displayed as a function of the beam length with the thickness as parameter for a given beam-contact gap of 2 nm and a beam width, W, of 8 nm. In order to optimize the geometry it is important to be able to control the beam thickness, i. e. the
198
S. Bengtsson et al.
nanotube diameter or in the case of bundles, the number of nanotubes. Still, exploiting the freedom of the design space and assuming a high Young’s modulus of 1 TPa, the carbon nanotube nanoelectromechanical switching will be more than one order of magnitude slower than a DRAM element at the same critical length (gate length and beam length). On the up side, the NEMS switch will consume orders of magnitude less power, mainly due to negligible off-state leakage current (1×10-5 µAµm-1 for CMOS DRAM and 1×10-9 µAµm-1 for the CNT switch). With a low actuation voltage design, the energy cost for a switching event can be in the aJ range10. 2.2. Varactors In the case of varactors one focus issue of the modeling we have employed is to correlate the effective Young’s modulus of the NEMS with the experimentally measurable pull-in voltage – the minimum voltage required to bring the two electrodes of the capacitor into contact, forming an electrical short-circuit. In a simplistic lumped capacitance model, depicted in Figure 3, it is possible to arrive at an analytical relationship for the pull-in voltage as a function of the geometrical dimensions of the actuated electrode and of its effective spring constant, k11. Such a model is not powerful enough to faithfully resolve the interplay between pull-in voltage, geometry and stiffness of an individual NEMS element to the extent that we can understand e. g. the impact of growth conditions on its experimentally observed mechanical behavior. Using the boundary element method (BEM) in simulations12 it is possible to obtain the deflection at a given geometry and biasing condition in an iterative fashion; the charge distribution is calculated first, then the resulting electric force, thereafter the deflection, after which the charge distribution (at the given voltage) is recalculated to start the next round of iteration. The outcome of such a simulation is displayed in Figure 4. In a design space delimited by estimations of
Fig. 4. This image shows the result of applying the BEM to calculate the nanofiber deflection for a given applied voltage.
Carbon-Based Nanoelectromechanical Devices
199
Fig. 5. Resulting relationship between pull-in voltage and Young’s modulus.
experimentally relevant geometries, voltages and Young’s moduli, the BEM simulations are run to generate a database of correlations between pull-in voltage and Young’s modulus; a measurement of the pull-in voltage can then be directly correlated to a value of the Young’s modulus for a given experimental condition, by referring to a generated relationship such as the one shown in Figure 5. 3. Device Fabrication and Characterization The technique of chemical vapor deposition (CVD) can produce CNTs and CNFs with a desired functional pattern. Any electronic function designed on a chip carries a cost in terms of its footprint, which makes it immediately favorable to design NEMS devices that extend upward, perpendicular to the chip surface. Depending on the processing details it is possible to obtain either individual free-standing vertically aligned CNFs (Figure 6) or
Fig. 6. An SEM picture of a pair of individually contacted CNFs separated by a distance below 500 nm.
200
S. Bengtsson et al.
Fig. 7. The forest of CNTs with a close up view to show the wiggly and porous structure. From Olofsson16 reprinted with permission from IOP Publishing.
forests of erect CNTs (Figure 7). In the case of CNFs the electric field of a plasma environment controls the growth directionality13, whereas it is the interaction between CNTs grown at a high enough density and rate which induces the vertical alignment in the forest configuration14. The CNFs have a complex internal structure with less order than a nanotube9, but one advantage is that individual vertically aligned nanostructures can be grown. It is not possible to achieve this with individual carbon nanotubes. 3.1. Individual Vertically Aligned Carbon Nanofibers Electron beam patterned nanoscale Ni dots have been employed to catalyze the growth of nanofibers with a diameter of 100 nm and lengths exceeding 1 m. Plasma CVD with a dc-plasma at a current of 20 mA was used to grow the fibers at 700 °C. The CNF pair of Figure 6 is an example of best achievement. The insulation between the individual fibers appears to be very good, with sub 0.1 nA current leakage up to 100 V. There is still need for further process optimization before we can fabricate single carbon nanofibers deterministically as a NEMS building block, but this development is on the level of optimization for specific processing tools, with the main hurdle being the choice of least
Fig. 8. Varactor electrodes made by carbon nanotube forests. Reprinted from Ek-Weis17 with permission from Professional Engineering Publishing.
Carbon-Based Nanoelectromechanical Devices
201
Fig. 9. a) Capacitance change as a function of voltage; b) and c) SEM images of the structure after exceeding pull-in. From Olofsson16 reprinted with permission from IOP Publishing.
interfering work-around to handle the discharging instabilities in the dc-plasma. Individual electrically addressable vertically aligned CNFs without mechanical degrees of freedom have been demonstrated15. 3.2. Carbon Nanotube Forest Using thermal CVD at 700 °C with Fe catalyst, forests of vertically aligned 135 m high multiwalled carbon nanotubes have been grown with a length of 200 m, a width of 4 m and a lateral separation of 10 m16. The areal density of nanotubes in the forest was estimated at 1010 nanotubes cm−2, and matching simulations to the measured actuation of these varactor electrodes yielded an effective Young’s modulus on the order of a few MPa, i. e. far below the TPa often attributed to individual nanotubes, thus allowing actuation to be achieved for relatively low applied voltages16. The very low effective Young’s modulus can be attributed to the highly porous and “wiggly” nature of the material (Figure 7). A varactor device designed using this material is displayed in Figure 8, showing the buttresses that need to be added to achieve sufficient mechanical rigidity. In Figure 9 the degree to which capacitance tuning can be achieved is shown along with the consequence of exceeding the pull-in voltage, as illustrated by two SEM images after the catastrophic event, which also alters the capacitance-voltage characteristic. The capacitance was determined by matching S parameter simulations to measurements, using an equivalent circuit which gave very good agreement with experimental observations17. 3.3. CMOS Compatibility It is a very attractive goal to be able to harness the advantages of carbon-based NEMS on a conventional CMOS electronics platform. Integration is a crucial issue when aiming for competitive system level performance for devices that incorporate and utilize NEMS.
202
S. Bengtsson et al.
Fig. 10. Deterioration of the on-state drain current of PMOS and NMOS transistors after three disparate growth processes. From Ghavanini18; reprinted with permission from ACS Publications.
However, finding a way to match the processing requirements to obtain good carbon nanostructures with the restrictions for processes and materials in CMOS production is not trivial. The outcome of exposing transistors fabricated in 130 nm bulk CMOS technology to carbon nanofiber growth conditions, i. e. elevated temperatures and a plasma environment, shows that transistors can survive such a treatment and even perform without immediate detrimental consequences. Comparing rf-plasma processing at 560 °C, thermal CVD at 610 °C and exposure to a dc-plasma at 500 °C, the last of these three nanofiber growth methods gives the least impact on transistor performance18, and in the case of the on-state drain current there is no discernible effect, as is shown in Figure 10. 4. Conclusion In theory, carbon-based nanoelectromechanical switches can provide a low static power technology with potential for high frequency operation. Components constituted by carbon nanotubes will display high mechanical stability and will also be highly stable with regards to temperature. However, the dependencies of device properties like these on the specific growth conditions need further investigations. Large scale reproducible
Carbon-Based Nanoelectromechanical Devices
203
and reliable manufacturing of integrated NEMS elements remains extremely challenging, where growth on top of CMOS as a back-end process further increases the complexity and limits the degrees of freedom by having to resolve the conflict between CMOS compatibility requirements and carbon nanostructure quality; this integration scenario is still however a future possibility albeit not obviously achievable. CNT-based NEMS switches demonstrate very low off-state leakage currents. Varactors can be realized with individual vertically aligned CNFs or with walls consisting of quasi-vertically aligned arrays of sparse CNTs. The latter are very porous, but behave mechanically as a cohesive unit with exceptional material properties. Acknowledgments The authors would like to acknowledge the external providers of financial support for this work, namely the sixth framework for research funding of the European Commission (contract 003673 CANEL and contract 028158 NANORF) and the Swedish Foundation for Strategic Research (SSF). References 1. A. P. Graham, G. S. Duesberg, R. Seidel, M. Liebau, E. Unger, F. Kreupl, and W. Honlein, Diamond and Related Materials 13, 1296-1300 (2004). 2. J. Robertson, Materials Today 10, 36-43 (2007). 3. P. Avouris and C. Jia, Materials Today 9, 46-54 (2006). 4. P. Chan, Y. Chai, Z. Min, and F. Yunyi, in The application of carbon nanotubes in CMOS integrated circuits, Piscataway, NJ, USA, 2008. 5. M. Haselman and S. Hauck, Proceedings of the IEEE 98, 11-38 (2010). 6. A. V. Melechko, R. Desikan, T. E. McKnight, K. L. Klein, and P. D. Rack, Journal of Physics D: Applied Physics 42, 193001 (28 pp.) (2009). 7. S. Reich, C. Thomsen, and J. Maultzsch, Carbon Nanotubes: Basic Concepts and Physical Properties, 1 ed. (Wiley-VCH, Berlin, 2004). 8. C. W. S. To, Finite Elements in Analysis and Design 42, 404-413 (2006). 9. A. V. Melechko, V. I. Merkulov, T. E. McKnight, M. A. Guillorn, K. L. Klein, D. H. Lowndes, and M. L. Simpson, Journal of Applied Physics 97, 39 (2005). 10. M. Y. A. Yousif, P. Lundgren, F. Ghavanini, P. Enoksson, and S. Bengtsson, Nanotechnology 19, 285204 (7 pp.) (2008). 11. M. Dequesnes, S. V. Rotkin, and N. R. Aluru, Nanotechnology 13, 120-131 (2002). 12. F. París and J. Cañas, Boundary element method : fundamentals and applications, Oxford Univ. Press (1997). 13. V. I. Merkulov, A. V. Melechko, M. A. Guillorn, M. L. Simpson, D. H. Lowndes, J. H. Whealton, and R. J. Raridon, Applied Physics Letters 80, 4816-4818 (2002). 14. G. Eres, A. A. Kinkhabwala, H. Cui, D. B. Geohegan, A. A. Puretzky, and D. H. Lowndes, Molecular beamcontrolled nucleation and growth of vertically aligned single-wall carbon nanotube arrays, The Journal of Physical Chemistry B 109, 16684-16694 (2005). 15. M. A. Guillorn, T. E. McKnight, A. Melechko, V. I. Merkulov, P. F. Britt, D. W. Austin, D. H. Lowndes, and M. L. Simpson, Journal of Applied Physics 91, 3824-8 (2002). 16. N. Olofsson, J. Ek-Weis, A. Eriksson, T. Idda, and E. E. B. Campbell, Nanotechnology 20 (2009).
204
S. Bengtsson et al.
17. J. Ek-Weis, A. Eriksson, T. Idda, N. Olofsson, and E. E. B. Campbell, Radio-frequency characterization of varactors based on carbon nanotube arrays, Journal Proceedings of the Institution of Mechanical Engineers, Part N: Journal of Nanoengineering and Nanosystems 222, 111-115 (2009). 18. F. A. Ghavanini, H. Le Poche, J. Berg, A. M. Saleem, M. S. Kabir, P. Lundgren, and P. Enoksson, Compatibility Assessment of CVD Growth of Carbon Nanofibers on Bulk CMOS Devices, Nano Letters 8, 2437-2441 (2008).
CHARGE PUDDLES AND EDGE EFFECT IN A GRAPHENE DEVICE AS STUDIED BY A SCANNING GATE MICROSCOPE J. CHAE, H. J. YANG, H. BAEK, J. HA, Y. KUK Department of Physics and Astronomy, Seoul National, University Gwanak-gu, Seoul, 151-747, Korea [email protected] S. Y. JUNG, Y. J. SONG, N. B. ZHITENEV, J. A. STROSCIO Center for Nanoscale Science and Technology, National Institute of Standards and Technology, 100 Bureau Drive, Gaithersburg, MD 20988, USA S. J. WOO, Y.-W. SON School of Computational Korea Institute of Advanced Studies, Dongdaemoon-gu, Seoul, 130-722, Korea Despite the recent progress in understanding the geometric structures of defects and edges in a graphene device (GD), how such defects and edges affect the transport properties of the device have not been clearly defined. In this study, the surface geometric structure of a GD was observed with an atomic force microscope (AFM) and the spatial variation of the transport current by the gating tip was measured with scanning gate microscopy (SGM). It was found that geometric corrugations, defects and edges directly influence the transport current. This observation is linked directly with a proposed scattering model based on macroscopic transport measurements. Keywords: Graphene Device; Scanning Gate Microscopy; Carrier Uniformity.
1. Introduction Graphene has been widely studied due to scientific interests and because of its possible application to high-speed devices1-4. Ever since the successful separation of graphene layers, many unique physical properties of graphene have been reported, including its linear dispersion relationship, relativistic fermionic behavior in the conduction and the valence bands, two-dimensional electron gas (2DEG) behavior, and back-scatteringless tunneling5-9. Many related theoretical predictions have been confirmed by macroscopic transport measurements2,3,10. Despite the successful explanation of the physical properties, the correlation between the geometric structures of defects such as those caused by corrugations, defects and edges and the carrier scattering at these defects is not fully understood. For example, the mobility of graphene measured in a suspended GD or epitaxially grown graphene is reportedly as high as 200,000 cm/V•s11-16. However, a GD on a SiO2 substrate has been measured in the range of 5,000~10,000 cm/V•s17-19. The difference is modeled by the carrier scattering by various defects on a GD on SiO2 substrate but not on a suspended GD. Thus far, structural studies have been 205
206
J. Chae et al.
conducted using microscopic tools such as electron microscopy and scanning tunneling microscopy16,20-25, while the transport properties have been measured macroscopically in a two- or four-terminal device with a back gate2. We report local carrier transport measurement results using SGM. SGM is a unique microscopic tool with which the local geometric and electronic structures and the transport property of an electronic device can be measured simultaneously26-34. An SGM uses a conducting tip to apply an electric field locally and measures the transport current through two or four contacts. It can use the same tip to measure the geometric structure in Atomic Force Microscopy (AFM) mode. The results of this experiment showed that the local geometrical defects indeed work as scattering centers in the transport measurement. 2. Transport Property of GD Even before any transport measurement of a GD were reported, Ando predicted the characteristics of the transport properties35; perfect graphene only with short-range scatters would show no conductivity dependence on the carrier density because the scattering rate would be divergent as the carrier density goes to zero near the Dirac point. However, the long-range scatters would be dominant in the low-density limit such that the device becomes insulating at the Dirac point. Earlier transport results2,3 reported that the conductivity is linearly dependent on the induced charge density by the back gate voltage, even in the range of ±1×1013 cm−2. This result suggests that the transport property of a GD is strongly dependent on the scattering mechanism by long-range scatters. With the progress of GD fabrication, a GD was processed as a suspended device in order to remove the substrate effects14,15. It was predicted that the carrier transport would be affected by scattering with charged impurities, short-range scatterers, mid-gap states, various phonon modes, surface corrugations, and defects in a GD17-19,36-40. In the case of the charged impurity potential at a high carrier density limit, the conductance is given by
σ
C
e2 n , h n imp
(1)
which is linearly dependent on the induced charge density17,38. The dimensionless constant C is ~10 to ~20. A recent experiment reported that the measured mobility is linearly dependent on the doping level of potassium38,41. Experimentally, the existence of charge puddle in graphene on the top of the SiO2 substrate was resolved using a scanning single-electron transistor microscope42. That experimental result was considered in the theoretical calculation of transport properties of GD near Dirac point43. Recently, another investigation resolved the charge impurity scattering centers using a spatial map in scanning tunneling spectroscopy (STS). In that study, the Dirac point was locally mapped from STS data and the scattering pattern was
Charge Puddles and Edge Effect in a Graphene Device
207
analyzed to determine the scattering centers. Good agreement was noted between the local Dirac point and a theoretical model of carrier scattering44. Another important scattering mechanism in a GD is scattering with a phonon. Raman spectroscopy results in the presence of an electric field showed G band damping due to scattering with a phonon40. The temperature-dependent resistivity of GD is well-fitted to the theoretical prediction considering the scattering with an acoustic and an optical phonon18. Scattering by geometric corrugations on a graphene surface is also an interesting subject. The corrugations can be formed by interaction with the substrate or by a thermal cycle during the annealing process45. From the theoretical calculations, electrons in the presence of a ripple are affected by the vector potential due to a strain field37,46. However, the role of corrugation in carrier transport remains not well understood. 3. Scanning Gate Microscope The SGM in this study was operated under an ultrahigh vacuum (UHV) at 20 K with liquid helium, at 84K with liquid nitrogen, and at room temperature. The sample can be annealed with current through a device or annealing of the whole chamber can be accomplished at an elevated temperature. An ultra-sharp AFM tip can be positioned over the GD using custom-made vacuum motors. During the SGM operation, the transport current is measured through the electrodes, as shown in Fig. 1A. As we place the
Fig. 1. (A) Schematic diagram of AFM operated at cryogenic temperature. This unit is kept inside a UHV tube that is inserted in a helium Dewar. A laser reflection from the back of a cantilever is detected by a quadrant photodiode. Piezoelectric motors are mounted to align the detection mechanism and the sample coarse movements. (B) CCD images of a GD with a long-range optical microscope installed inside UHV AFM chamber. (C) Operational principle of SGM. As an AFM cantilever scans a line, the controlling computer remembers the topography. After the scan a tip bias is applied to the tip and follows the topography, such that the tip follows the same distance away from the sample.
208
J. Chae et al.
conducting AFM over the area of interest in a GD, the transport current varies as we move the device. A typical total conductance value without the tip gate is ~ 100 S whereas the variation in the conductance with the tip gate is ~ 1 S. A custom-made UHV AFM operating at cryogenic temperatures was used to perform the SGM experiment, as shown in Fig. 1B47. The AFM head has a rigid three-column structure inside the UHV chamber, and the chamber is immersed in a liquid-helium Dewar and cooled by the helium exchange gas (at nearly one atmospheric pressure) to ensure optimal thermal and vibrational stability. The radiation through the main probe chamber is effectively cut off by internal baffles and the objective lens of the CCD optics is installed for a long-range optical microscope just above the AFM head. To achieve the maximum numerical aperture, we mounted an objective lens with a diameter of 25 mm and a focal length of 60 mm on top of the head. We drilled a small hole (4 mm diameters) in the lens along its optical axis to allow for the unhindered piezomotor-driven motion of the optical fibers. The intermediate image of the tip and the sample formed near the top of the main chamber is refocused onto the CCD by a telescopic microscope lens. Fig. 1C shows a long-range optical microscope image on the sample surface and the cantilever. Vibration isolation is achieved by anchoring the chamber on a granite table. Three air table legs support the granite table, at 500 kg, and the cut off frequency is below ~3Hz. A helium Dewar is attached to three additional air dampers to isolate the mechanical vibration from the Dewar. The system is installed in a sound-proof room that is separate from the control and measurement circuits. The noise level from mechanical parts is much less than 100dB. A laser light bounced off the back of a commercial cantilever is detected by a quadrant photo diode to detect the z movement of the probe. The head has four inertial piezomotors: a one-dimensional (1D) Z motor48 (10 mm span) for a coarse approach to the sample, and three planar XY motors (2 mm x 2 mm span each) for fine horizontal positioning of the sample and to adjust the laser lens assembly and the photodiode. A pan-type linear walker with six independent shear piezo-ceramic feet is used for coarse movement of the sample to the tip. As a white light source (the cleaved end of the multi-mode fiber) illuminates the sample surface at the angle , we can estimate the distance between the tip and the sample to be ten times the distance between the cantilever and its shadow laid on the sample, achieving a safe and fast coarse approach down to a tip-sample distance of ~ 5 µm. 4. Sample Preparation Sub-headings The GDs used in this study were fabricated by fabrication processes similar to those reported earlier4. Graphene flakes were mechanically exfoliated from either highly oriented pyrolitic graphite (HOPG) or natural graphite. Graphite was tapped onto scotch tape and then peeled off several times until it appeared shiny. Subsequently, this graphene tape was transferred by finger onto the top of thermally grown 300-nm thick SiO2 on a highly doped Si substrate that was used as back gate later. The thickness of the graphene flake was confirmed by micro-Raman spectroscopy. A single layer of graphene
Charge Puddles and Edge Effect in a Graphene Device
209
was confirmed via image contrast of an optical microscope. Source-drain metal (3 nm Cr / 30 nm Au) contacts were defined by conventional electron beam lithography and a lift-off process. Figure 2A shows an optical image of a fabricated GD with metallic contacts. These electrodes were used as markers to position the sample with the tip later, as shown in Fig. 1C. After the GD was mounted in a cryogenic AFM, a macroscopic transport measurement was performed to check the device characteristics, including the position of the Dirac point, electron and the hole mobility, as shown in Fig. 2B. Red and green lines represent data of a different sweep direction to note the hysteresis of the device. As the GDs were fabricated in an ambient environment and transferred to the AFM chamber, the macroscopic transport properties changed in the UHV chamber, as expected49. Figures 2C and D demonstrate the transport current as a function of the back gate bias voltage before the GD was inserted into a vacuum (2C) and in the vacuum (2D). Under ambient pressure, the Dirac point of the GD exceeded +20 V, indicating a holedoped device by water molecule adsorption on graphene. After the GD was annealed at 380 K for 6 hours, the Dirac point moved to ~0 V, suggesting that the adsorbates were desorbed under UHV. The mobility of the sample was also increased when this process was repeated.
Fig. 2. (A) An optical microscope image of a fabricated GD. A faint dark object is a single- layer graphene and bright parts are contact electrodes. (B) A typical graphene resistance versus back gate bias voltage after loading into AFM and several annealing process. Red data was taken as increasing the bias and the green curve was obtained as ramping down. (C)-(D) Macroscopic transport measurement data before inserting into AFM chamber (C) and after annealing at 380K for 6 hours (D).
210
J. Chae et al.
5. Results and Discussion The topography of a graphene layer was measured using a scanning tunneling microscope (STM). In monolayer graphene, a honeycomb-shaped STM topography is observed. In double-layer graphene, every other carbon atom appears bright, resulting in a hexagonal topography due to the Bernal stacking of the double layers. Figure 3A shows an STM image of an epitaxially grown graphene layer on a Si-terminated SiC surface50, showing the honeycomb-shaped atomic details. Imaging of the atomic details of a GD on SiO2 was not possible with STM without an x-y motor or by the optical microscope in this chamber. In the SGM chamber, where x-y-z motors were equipped, it was possible to obtain a resolution of ~ 1 nm with a cantilever-based operation. Figure 3B shows a STS result on the graphene layer. It has a hump just below the Fermi level, as explained by phonon-mediated tunneling near the Dirac point25. After a GD was installed in a cryogenic AFM-SGM chamber, it was imaged in AFM mode. Figure 4A shows a contact mode AFM image of a GD, where a SiO2-coated Si tip was used. The GD appears to have a lighter color in the middle. The corrugation on the SiO2 surface is known to be ~1 nm with an average corrugation width of ~50-60 nm51. Similar corrugation was observed on the GD, and the characteristic features are very similar on the GD and a SiO2 substrate. Figure 4B shows the AFM topography of a GD with a higher resolution. Again, the observed corrugation depth is very similar to that on a SiO2 surface and the average separation among those corrugations is ~50-60 nm, as reported on a bare SiO2 surface. These data suggest if there is a scattering mechanism due to corrugation, it can be modeled by this topography. Unlike earlier results36, we could not find self-similarity in the length scale in our data.
Fig. 3. (A) STM topography of a monolayer graphene on a Si-terminated SiC. (B) STS data on the surface of the monolayer graphene.
Charge Puddles and Edge Effect in a Graphene Device
211
Fig. 4. (A) AFM topography of a GD measured in contact mode. (B) An AFM topography in non-contact mode. Maximum vertical corrugation is ~ 2 nm.
Fig. 5. (A) An AFM topography of 800 x 800 nm area on a GD. (B)-(L) SGM data on the same area at tip gating bias of −2.5 V, −2.0 V, −1.5 V, −1.0 V, −0.5 V, 0 V, 0.5 V, 1.0 V, 1.5 V, 2.0 V and 2.5 V respectively. (M) Color-coded scale bar of conductance change for SGM data.
After the corrugation was confirmed by AFM topography, SGM experiments were conducted. Figure 5A depicts a 800 x 800 nm2 AFM topography of a GD and Figs. 5B-L show SGM micrographs from a tip bias voltage range of –2.5 to +2.5 in steps of 0.5 V. The Dirac point of this GD was –24 V, as set by the back gate bias. All of the SGM measurements were taken at Dirac point by back-gating. Corrugations on graphene
212
J. Chae et al.
surface are seen in topography image with lateral dimension of ~100 nm. The spatial variations of conductance change become more prominent as the absolute value of the tip gate bias increases in either negative or positive polarity. Negative (positive) polarity of electric field from the tip is related to the scattering dynamics of electron (hole) carriers. The small features in the SGM data are tens of nanometers, revealing good agreement with the AFM observation. There is strong correlation between the SGM signal and the topographic corrugations; the peaks appear at the bottom of the valley in topography. The range of high conductance region with hole carrier is extended more than that with electron carrier at the same electric field strength. This difference is due to the structure of charge puddle. The correlation between topographic data and SGM data implies that the charge puddles exist at the bottom of each ripple. The difference features in electron and hole carriers indicates that electrons are more locally confined at the bottom of the ripple induced by the interaction with substrate and hole carrier screens around confined electrons. Unlike scattering by a geometrical corrugation, scattering caused by crystalline defects are slightly different. Figure 6A shows the topographic image with defect and Figs. 6B-F show the tip bias dependence of the SGM micrographs near defects with tip gating bias of −2.0 V (B), −4.0 V (C), 0 V (D), 2.0 V (E) and 4.0 V (F) respectively. The nature of this defect is not clear, but this area appears to be elastically deformed. The SGM signal indicates strong carrier scattering even at a high tip bias voltage52 compared to scattering by geometrical corrugations in which the averaging effect by tip bias conceal the spatial resolution with tip gating voltage of higher than ±3 V. It is worth to note that the conductance change by tip gating is dramatically high compared to scattering by geometrical corrugations. Color-coded scale bar for SGM signal for defect is shown in Fig. 6G which is an order of magnitude higher than that for geometrical corrugation as shown in Fig. 5M. This result indicates that the defect can affect the transport properties in a dramatic way and has to be considered carefully.
Fig. 6. (A) An AFM topography of geometrical defect. (B)-(F) Corresponding SGM signals with tip gating bias of −2 V, −4 V, 0 V, 2 V and 4 V respectively. (G) Color-coded scale bar of conductance change for SGM data.
Charge Puddles and Edge Effect in a Graphene Device
213
Fig. 7. SGM micrograph of a GD in which the Dirac point is -20 V of the back gate voltage with tip gate bias of −1 V (A) and −2 V (B). Strong conductance enhancement at the edge was observed with −2 V of tip bias which is more close to Dirac point.
When the SGM signal was measured at edges of a GD, a strong enhancement of the SGM signal was observed. Figures 7A and B show SGM results obtained by contact mode AFM with SiO2-coated Si tip. The Dirac point of this sample was –20 V according to the back gate bias. As the tip gate bias was decreased to –2 V, Fig. 7B, i.e., as the tip gate approaches the Dirac point, the edge transport signal increases by as much as 10 times that at the middle of the GD. That is due to charge accumulation at the edge of a GD by electrostatic force53. The edge state conducting channel can be opened and contributed to electron transport like in a lightning rod. More experiment about edge effect with gate-controlled GD showed that the conductance enhancement at the edge was not observed with back gate of Dirac point (in low carrier density limit), however, the conductance enhancement at edge were observed again in electron or hole charging regime. Theoretical calculation with SGM simulation confirmed this edge effect by the edge channel opening52. Based on this result, it becomes necessary to re-interpret a considerable amount of existing macroscopic data. In conclusion, this SGM study shows that geometric ripples, defects and edges directly influence the transport current. This observation is directly linked with the proposed scattering model based on macroscopic transport measurements, and the existing results must be re-examined. This work was supported by National Research Foundation of Korea (334820090042).
214
J. Chae et al.
References 1. K.S. Novoselov, A.K. Geim, S.V. Morozov, D. Jiang, Y. Zhang, S.V. Dubonos, et al., Electric Field Effect in Atomically Thin Carbon Films, Science. 306 (2004) 666-669. 2. K.S. Novoselov, A.K. Geim, S.V. Morozov, D. Jiang, M.I. Katsnelson, I.V. Grigorieva, et al., Two-dimensional gas of massless Dirac fermions in graphene, Nature. 438 (2005) 197-200. 3. Y. Zhang, Y. Tan, H.L. Stormer, P. Kim, Experimental observation of the quantum Hall effect and Berry’s phase in graphene, Nature. 438 (2005) 201-204. 4. K.S. Novoselov, D. Jiang, F. Schedin, T.J. Booth, V.V. Khotkevich, S.V. Morozov, et al., Two-dimensional atomic crystals, Proceedings of the National Academy of Sciences of the United States of America. 102 (2005) 10451-10453. 5. A.H. Castro Neto, F. Guinea, N.M.R. Peres, Drawing conclusions from graphene, Physics World. 19 (2006) 33-37. 6. V.P. Gusynin, S.G. Sharapov, Unconventional Integer Quantum Hall Effect in Graphene, Phys. Rev. Lett. 95 (2005) 146801. 7. N.M.R. Peres, F. Guinea, A.H. Castro Neto, Electronic properties of disordered twodimensional carbon, Phys. Rev. B. 73 (2006) 125411. 8. M.I. Katsnelson, K.S. Novoselov, A.K. Geim, Chiral tunnelling and the Klein paradox in graphene, Nat Phys. 2 (2006) 620-625. 9. M. Katsnelson, K. Novoselov, Graphene: New bridge between condensed matter physics and quantum electrodynamics, Solid State Communications. 143 (2007) 3-13. 10. A.F. Young, P. Kim, Quantum interference and Klein tunnelling in graphene heterojunctions, Nat Phys. 5 (2009) 222-226. 11. K. Bolotin, K. Sikes, Z. Jiang, M. Klima, G. Fudenberg, J. Hone, et al., Ultrahigh electron mobility in suspended graphene, Solid State Communications. 146 (2008) 351-355. 12. X. Du, I. Skachko, A. Barker, E.Y. Andrei, Approaching ballistic transport in suspended graphene, Nat Nano. 3 (2008) 491-495. 13. M. Orlita, C. Faugeras, P. Plochocka, P. Neugebauer, G. Martinez, D.K. Maude, et al., Approaching the Dirac Point in High-Mobility Multilayer Epitaxial Graphene, Phys. Rev. Lett. 101 (2008) 267601. 14. K.I. Bolotin, F. Ghahari, M.D. Shulman, H.L. Stormer, P. Kim, Observation of the fractional quantum Hall effect in graphene, Nature. 462 (2009) 196-199. 15. X. Du, I. Skachko, F. Duerr, A. Luican, E.Y. Andrei, Fractional quantum Hall effect and insulating phase of Dirac electrons in graphene, Nature. 462 (2009) 192-195. 16. D.L. Miller, K.D. Kubista, G.M. Rutter, M. Ruan, W.A. de Heer, P.N. First, et al., Observing the Quantization of Zero Mass Carriers in Graphene, Science. 324 (2009) 924-927. 17. Y. Tan, Y. Zhang, K. Bolotin, Y. Zhao, S. Adam, E.H. Hwang, et al., Measurement of Scattering Rate and Minimum Conductivity in Graphene, Phys. Rev. Lett. 99 (2007) 246803. 18. J. Chen, C. Jang, S. Xiao, M. Ishigami, M.S. Fuhrer, Intrinsic and extrinsic performance limits of graphene devices on SiO2, Nat Nano. 3 (2008) 206-209. 19. J. Chen, C. Jang, M. Ishigami, S. Xiao, W. Cullen, E. Williams, et al., Diffusive charge transport in graphene on SiO2, Solid State Communications. 149 (2009) 1080-1086. 20. F. Wang, Y. Zhang, C. Tian, C. Girit, A. Zettl, M. Crommie, et al., Gate-Variable Optical Transitions in Graphene, Science. 320 (2008) 206-209. 21. Z. Liu, K. Suenaga, P.J.F. Harris, S. Iijima, Open and Closed Edges of Graphene Layers, Phys. Rev. Lett. 102 (2009) 015501. 22. X. Jia, M. Hofmann, V. Meunier, B.G. Sumpter, J. Campos-Delgado, J.M. Romo-Herrera, et al., Controlled Formation of Sharp Zigzag and Armchair Edges in Graphitic Nanoribbons, Science. 323 (2009) 1701-1705.
Charge Puddles and Edge Effect in a Graphene Device
215
23. C.O. Girit, J.C. Meyer, R. Erni, M.D. Rossell, C. Kisielowski, L. Yang, et al., Graphene at the Edge: Stability and Dynamics, Science. 323 (2009) 1705-1708. 24. G.M. Rutter, J.N. Crain, N.P. Guisinger, T. Li, P.N. First, J.A. Stroscio, Scattering and Interference in Epitaxial Graphene, Science. 317 (2007) 219-222. 25. Y. Zhang, V.W. Brar, F. Wang, C. Girit, Y. Yayon, M. Panlasigui, et al., Giant phononinduced conductance in scanning tunnelling spectroscopy of gate-tunable graphene, Nat Phys. 4 (2008) 627-630. 26. E.J. Heller, K.E. Aidala, B.J. LeRoy, A.C. Bleszynski, A. Kalben, R.M. Westervelt, et al., Thermal Averages in a Quantum Point Contact with a Single Coherent Wave Packet, Nano Letters. 5 (2005) 1285-1292. 27. G. Metalidis, P. Bruno, Green’s function technique for studying electron flow in twodimensional mesoscopic samples, Phys. Rev. B. 72 (2005) 235304. 28. A. Freyn, I. Kleftogiannis, J. Pichard, Scanning Gate Microscopy of a Nanostructure Where Electrons Interact, Phys. Rev. Lett. 100 (2008) 226802. 29. M.G. Pala, B. Hackens, F. Martins, H. Sellier, V. Bayot, S. Huant, et al., Local density of states in mesoscopic samples from scanning gate microscopy, Phys. Rev. B. 77 (2008) 125310. 30. J.W.P. Hsu, N.G. Weimann, M.J. Manfra, K.W. West, D.V. Lang, F.F. Schrey, et al., Effect of dislocations on local transconductance in AlGaN/GaN heterostructures as imaged by scanning gate microscopy, Appl. Phys. Lett. 83 (2003) 4559. 31. B. Hackens, F. Martins, T. Ouisse, H. Sellier, S. Bollaert, X. Wallart, et al., Imaging and controlling electron transport inside a quantum ring, Nat Phys. 2 (2006) 826-830. 32. M.A. Topinka, B.J. LeRoy, R.M. Westervelt, S.E.J. Shaw, R. Fleischmann, E.J. Heller, et al., Coherent branched flow in a two-dimensional electron gas, Nature. 410 (2001) 183-186. 33. K.E. Aidala, R.E. Parrott, E.J. Heller, R.M. Westervelt, Imaging electrons in a magnetic field, Cond-Mat/0603035. (2006). 34. M.P. Jura, M.A. Topinka, L. Urban, A. Yazdani, H. Shtrikman, L.N. Pfeiffer, et al., Unexpected features of branched flow through high-mobility two-dimensional electron gases, Nat Phys. 3 (2007) 841-845. 35. N. Shon, T. Ando, Quantum Transport in Two-Dimensional Graphite System, J. Phys. Soc. Jpn. 67 (1998) 2421-2429. 36. M. Ishigami, J.H. Chen, W.G. Cullen, M.S. Fuhrer, E.D. Williams, Atomic Structure of Graphene on SiO2, Nano Letters. 7 (2007) 1643-1648. 37. M. Katsnelson, A. Geim, Electron scattering on microscopic corrugations in graphene, Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences. 366 (2008) 195-204. 38. J. Chen, C. Jang, S. Adam, M.S. Fuhrer, E.D. Williams, M. Ishigami, Charged-impurity scattering in graphene, Nat Phys. 4 (2008) 377-381. 39. L.A. Ponomarenko, R. Yang, T.M. Mohiuddin, M.I. Katsnelson, K.S. Novoselov, S.V. Morozov, et al., Effect of a High- kappa Environment on Charge Carrier Mobility in Graphene, Phys. Rev. Lett. 102 (2009) 206603. 40. J. Yan, Y. Zhang, P. Kim, A. Pinczuk, Electric Field Effect Tuning of Electron-Phonon Coupling in Graphene, Phys. Rev. Lett. 98 (2007) 166802. 41. C. Jang, S. Adam, J. Chen, E.D. Williams, S. Das Sarma, M.S. Fuhrer, Tuning the Effective Fine Structure Constant in Graphene: Opposing Effects of Dielectric Screening on Short- and Long-Range Potential Scattering, Phys. Rev. Lett. 101 (2008) 146805. 42. J. Martin, N. Akerman, G. Ulbricht, T. Lohmann, J.H. Smet, K. von Klitzing, et al., Observation of electron-hole puddles in graphene using a scanning single-electron transistor, Nat Phys. 4 (2008) 144-148.
216
J. Chae et al.
43. E. Rossi, S. Adam, S. Das Sarma, Effective medium theory for disordered two-dimensional graphene, Phys. Rev. B. 79 (2009) 245423. 44. Y. Zhang, V.W. Brar, C. Girit, A. Zettl, M.F. Crommie, Origin of spatial charge inhomogeneity in graphene, Nat Phys. 5 (2009) 722-726. 45. C. Chen, W. Bao, J. Theiss, C. Dames, C.N. Lau, S.B. Cronin, Raman Spectroscopy of Ripple Formation in Suspended Graphene, Nano Letters. 9 (2009) 4172-4176. 46. S.V. Morozov, K.S. Novoselov, M.I. Katsnelson, F. Schedin, L.A. Ponomarenko, D. Jiang, et al., Strong Suppression of Weak Localization in Graphene, Phys. Rev. Lett. 97 (2006) 016801. 47. J. Lee, J. Chae, C.K. Kim, H. Kim, S. Oh, Y. Kuk, Versatile low-temperature atomic force microscope with in situ piezomotor controls, charge-coupled device vision, and tip-gated transport measurement capability, Rev. Sci. Instrum. 76 (2005) 093701. 48. S.H. Pan, E.W. Hudson, J.C. Davis, [sup 3]He refrigerator based very low temperature scanning tunneling microscope, Rev. Sci. Instrum. 70 (1999) 1459. 49. T. Lohmann, K. von Klitzing, J.H. Smet, Four-Terminal Magneto-Transport in Graphene p-n Junctions Created by Spatially Selective Doping, Nano Letters. 9 (2009) 1973-1979. 50. H. Yang, A.J. Mayne, M. Boucherit, G. Comtet, G. Dujardin, Y. Kuk, Quantum Interference Channeling at Graphene Edges, Nano Letters. 10 (2010) 943-947. 51. X. Blasco, D. Hill, M. Porti, M. Nafria, X. Aymerich, Topographic characterization of AFMgrown SiO_{2} on Si, Nanotechnology. 12 (2001) 110. 52. J. Chae, S.Y. Jung, S.J. Woo, H.J. Yang, H. Baek, J. Ha1, Y.J. Song, Y.-W. Son, N.B. Zhitenev, J.A. Stroscio, Y. Kuk, to be published (2010). 53. P.G. Silvestrov, K.B. Efetov, Charge accumulation at the boundaries of a graphene strip induced by a gate voltage: Electrostatic approach, Phys. Rev. B. 77 (2008) 155436.
WIDE BAND GAP TECHNOLOGY FOR HIGH POWER AND UV PHOTONICS
This page intentionally left blank
NOVEL APPROACHES TO MICROWAVE SWITCHING DEVICES USING NITRIDE TECHNOLOGY G. SIMIN*, J. WANG, B. KHAN Electrical Engineering, University of South Carolina, Columbia, SC, 29208, USA *[email protected] J. YANG, A. SATTU, R. GASKA Sensor Electronic Technology Inc., Columbia, SC, 29209, USA M. SHUR Electrical, Computer and System Eng., Rensselaer Polytechnic Institute, Troy, N.Y, USA III-Nitride heterostructure field-effect transistors (HFETs) demonstrated a new paradigm in microwave switching and control applications due to unique combination of extremely low channel resistance (leading to low loss), very high RF power, low off-state capacitance, broad range of operating temperatures, chemical inertness and robustness. The paper reviews novel approaches and recent advances in III-Nitride technology for RF switching devices leading to higher operating frequencies and even lower insertion loss. Keywords: Gallium nitride; Heterostructure Field-Effect transistors; Microwave Switches.
1. Introduction Importance of RF switching and control devices in modern microwave systems continuously increases along with increase in their complexity and functionality. Software Defined Radio (SDR) systems, satellite communications, broad-band and multiband mobile communications, and beam steering radars all require highly reliable, fast, low-loss RF switching components. Currently, pin-diodes, GaAs MESFETs or HEMTs and RF MEMS are commonly used as RF switching and control components. All of these devices have significant performance limitations arising from the fundamental materials properties and/or device design such as significant forward currents and turn-off times in pin-diodes, high control voltages, self-actuation, low RF powers and reliability issues in RF MEMS, low RF powers in GaAs-based MESFETs and HEMTs. AlInGaN based heterostructure field-effect transistors (HFETs) enable a new paradigm in the RF switch design. In these devices, the 2D electron gas having a record high density forms a metal-like conducting plate, which can be easily turned on and off by moderate gate control voltages. These features, in combination with extremely high breakdown voltages, chemical inertness and planar structure form a platform for excellent switching devices. The insulated gate device design (MOSHFET) further develops these 219
220
G. Simin et al.
features into reliable and robust superior RF switch with ultra-low control powers, high linearity and high RF powers. RF switches using III-Nitride MOSHFETs can operate with 10 - 100 times higher power of RF signals as compared to the available RF MEMS and GaAs HEMTs. They have two-three orders of magnitude lower power consumption and switching times than those of pin-switches. They also have high reliability and thermal stability with operating temperatures ranging from cryogenic temperatures up to 300°C or even higher. These features make the III-Nitride technology to be the best candidate to replace most of the existing RF switching components. The achieved results confirm great potential of III-N MOSHFETs as ultimate RF switching devices and MMICs 1-6 2. Key III-N Switch Performance Parameters and Limitations In a typical series transistor switch configuration, the source and drain electrodes are connected to the RF line input and output correspondingly. The gate electrode is connected to a control voltage supply through a blocking resistor. Fig. 1 shows the equivalent circuit of the series FET RF switch, including the variable, gate-voltage controlled channel resistance, parasitic device capacitances and inductors associated with the bonding wires. In the “ON” state, the HFET or MOSHFET switch gate bias is zero or positive, the CDS RCh(VG) LW LW channel resistance RON is low that ensures low-loss input-output transmission. The drain-source capacitance is shunted by the RF-In RF-Out C C GD Ω Ω GS RON, therefore the transmission is almost (50 ) (50 ) VG frequency independent. The value of the RON can be estimated as RON = 2RC + RGS+RGD + Fig. 1. Equivalent circuit of a series FET switch including device parasitic capacitances and RG, where RC is the contact resistance, RGS = RSH×LGS/W and RGD = RSH×LGD/W, RSH mounting wire inductances (after7). =1/(qNSµn) is the layer sheet resistance, W is the device width, NS is the channel sheet electron density, µn is the channel electron mobility, LGS, LGD are the source-gate and gate-drain spacing, and RG is the voltage dependent resistance of the channel under the gate. At zero or positive gate bias RG ≈ RSH×LG/W, where LG is the gate length. For a typical MOSHFET, RSH ≈ 300 Ω/sq, LG = 1 µm, LGS ≈ LGD ≈ 1.5 µm, the contact resistance, RC ≈ 0.5 Ω×mm, RON ≈ 2.2 Ω×mm. The insertion loss of a series resistor RON connected into a transmission line with the characteristic impedance Z0 (Z0 = 50 Ω typically), assuming RON